1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * NET An implementation of the SOCKET network access protocol. 4 * 5 * Version: @(#)socket.c 1.1.93 18/02/95 6 * 7 * Authors: Orest Zborowski, <obz@Kodak.COM> 8 * Ross Biro 9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 10 * 11 * Fixes: 12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in 13 * shutdown() 14 * Alan Cox : verify_area() fixes 15 * Alan Cox : Removed DDI 16 * Jonathan Kamens : SOCK_DGRAM reconnect bug 17 * Alan Cox : Moved a load of checks to the very 18 * top level. 19 * Alan Cox : Move address structures to/from user 20 * mode above the protocol layers. 21 * Rob Janssen : Allow 0 length sends. 22 * Alan Cox : Asynchronous I/O support (cribbed from the 23 * tty drivers). 24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style) 25 * Jeff Uphoff : Made max number of sockets command-line 26 * configurable. 27 * Matti Aarnio : Made the number of sockets dynamic, 28 * to be allocated when needed, and mr. 29 * Uphoff's max is used as max to be 30 * allowed to allocate. 31 * Linus : Argh. removed all the socket allocation 32 * altogether: it's in the inode now. 33 * Alan Cox : Made sock_alloc()/sock_release() public 34 * for NetROM and future kernel nfsd type 35 * stuff. 36 * Alan Cox : sendmsg/recvmsg basics. 37 * Tom Dyas : Export net symbols. 38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n". 39 * Alan Cox : Added thread locking to sys_* calls 40 * for sockets. May have errors at the 41 * moment. 42 * Kevin Buhr : Fixed the dumb errors in the above. 43 * Andi Kleen : Some small cleanups, optimizations, 44 * and fixed a copy_from_user() bug. 45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0) 46 * Tigran Aivazian : Made listen(2) backlog sanity checks 47 * protocol-independent 48 * 49 * This module is effectively the top level interface to the BSD socket 50 * paradigm. 51 * 52 * Based upon Swansea University Computer Society NET3.039 53 */ 54 55 #include <linux/bpf-cgroup.h> 56 #include <linux/ethtool.h> 57 #include <linux/mm.h> 58 #include <linux/socket.h> 59 #include <linux/file.h> 60 #include <linux/splice.h> 61 #include <linux/net.h> 62 #include <linux/interrupt.h> 63 #include <linux/thread_info.h> 64 #include <linux/rcupdate.h> 65 #include <linux/netdevice.h> 66 #include <linux/proc_fs.h> 67 #include <linux/seq_file.h> 68 #include <linux/mutex.h> 69 #include <linux/if_bridge.h> 70 #include <linux/if_vlan.h> 71 #include <linux/ptp_classify.h> 72 #include <linux/init.h> 73 #include <linux/poll.h> 74 #include <linux/cache.h> 75 #include <linux/module.h> 76 #include <linux/highmem.h> 77 #include <linux/mount.h> 78 #include <linux/pseudo_fs.h> 79 #include <linux/security.h> 80 #include <linux/syscalls.h> 81 #include <linux/compat.h> 82 #include <linux/kmod.h> 83 #include <linux/audit.h> 84 #include <linux/wireless.h> 85 #include <linux/nsproxy.h> 86 #include <linux/magic.h> 87 #include <linux/slab.h> 88 #include <linux/xattr.h> 89 #include <linux/nospec.h> 90 #include <linux/indirect_call_wrapper.h> 91 #include <linux/io_uring/net.h> 92 93 #include <linux/uaccess.h> 94 #include <asm/unistd.h> 95 96 #include <net/compat.h> 97 #include <net/wext.h> 98 #include <net/cls_cgroup.h> 99 100 #include <net/sock.h> 101 #include <linux/netfilter.h> 102 103 #include <linux/if_tun.h> 104 #include <linux/ipv6_route.h> 105 #include <linux/route.h> 106 #include <linux/termios.h> 107 #include <linux/sockios.h> 108 #include <net/busy_poll.h> 109 #include <linux/errqueue.h> 110 #include <linux/ptp_clock_kernel.h> 111 #include <trace/events/sock.h> 112 113 #ifdef CONFIG_NET_RX_BUSY_POLL 114 unsigned int sysctl_net_busy_read __read_mostly; 115 unsigned int sysctl_net_busy_poll __read_mostly; 116 #endif 117 118 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to); 119 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from); 120 static int sock_mmap(struct file *file, struct vm_area_struct *vma); 121 122 static int sock_close(struct inode *inode, struct file *file); 123 static __poll_t sock_poll(struct file *file, 124 struct poll_table_struct *wait); 125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 126 #ifdef CONFIG_COMPAT 127 static long compat_sock_ioctl(struct file *file, 128 unsigned int cmd, unsigned long arg); 129 #endif 130 static int sock_fasync(int fd, struct file *filp, int on); 131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 132 struct pipe_inode_info *pipe, size_t len, 133 unsigned int flags); 134 static void sock_splice_eof(struct file *file); 135 136 #ifdef CONFIG_PROC_FS 137 static void sock_show_fdinfo(struct seq_file *m, struct file *f) 138 { 139 struct socket *sock = f->private_data; 140 const struct proto_ops *ops = READ_ONCE(sock->ops); 141 142 if (ops->show_fdinfo) 143 ops->show_fdinfo(m, sock); 144 } 145 #else 146 #define sock_show_fdinfo NULL 147 #endif 148 149 /* 150 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear 151 * in the operation structures but are done directly via the socketcall() multiplexor. 152 */ 153 154 static const struct file_operations socket_file_ops = { 155 .owner = THIS_MODULE, 156 .llseek = no_llseek, 157 .read_iter = sock_read_iter, 158 .write_iter = sock_write_iter, 159 .poll = sock_poll, 160 .unlocked_ioctl = sock_ioctl, 161 #ifdef CONFIG_COMPAT 162 .compat_ioctl = compat_sock_ioctl, 163 #endif 164 .uring_cmd = io_uring_cmd_sock, 165 .mmap = sock_mmap, 166 .release = sock_close, 167 .fasync = sock_fasync, 168 .splice_write = splice_to_socket, 169 .splice_read = sock_splice_read, 170 .splice_eof = sock_splice_eof, 171 .show_fdinfo = sock_show_fdinfo, 172 }; 173 174 static const char * const pf_family_names[] = { 175 [PF_UNSPEC] = "PF_UNSPEC", 176 [PF_UNIX] = "PF_UNIX/PF_LOCAL", 177 [PF_INET] = "PF_INET", 178 [PF_AX25] = "PF_AX25", 179 [PF_IPX] = "PF_IPX", 180 [PF_APPLETALK] = "PF_APPLETALK", 181 [PF_NETROM] = "PF_NETROM", 182 [PF_BRIDGE] = "PF_BRIDGE", 183 [PF_ATMPVC] = "PF_ATMPVC", 184 [PF_X25] = "PF_X25", 185 [PF_INET6] = "PF_INET6", 186 [PF_ROSE] = "PF_ROSE", 187 [PF_DECnet] = "PF_DECnet", 188 [PF_NETBEUI] = "PF_NETBEUI", 189 [PF_SECURITY] = "PF_SECURITY", 190 [PF_KEY] = "PF_KEY", 191 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE", 192 [PF_PACKET] = "PF_PACKET", 193 [PF_ASH] = "PF_ASH", 194 [PF_ECONET] = "PF_ECONET", 195 [PF_ATMSVC] = "PF_ATMSVC", 196 [PF_RDS] = "PF_RDS", 197 [PF_SNA] = "PF_SNA", 198 [PF_IRDA] = "PF_IRDA", 199 [PF_PPPOX] = "PF_PPPOX", 200 [PF_WANPIPE] = "PF_WANPIPE", 201 [PF_LLC] = "PF_LLC", 202 [PF_IB] = "PF_IB", 203 [PF_MPLS] = "PF_MPLS", 204 [PF_CAN] = "PF_CAN", 205 [PF_TIPC] = "PF_TIPC", 206 [PF_BLUETOOTH] = "PF_BLUETOOTH", 207 [PF_IUCV] = "PF_IUCV", 208 [PF_RXRPC] = "PF_RXRPC", 209 [PF_ISDN] = "PF_ISDN", 210 [PF_PHONET] = "PF_PHONET", 211 [PF_IEEE802154] = "PF_IEEE802154", 212 [PF_CAIF] = "PF_CAIF", 213 [PF_ALG] = "PF_ALG", 214 [PF_NFC] = "PF_NFC", 215 [PF_VSOCK] = "PF_VSOCK", 216 [PF_KCM] = "PF_KCM", 217 [PF_QIPCRTR] = "PF_QIPCRTR", 218 [PF_SMC] = "PF_SMC", 219 [PF_XDP] = "PF_XDP", 220 [PF_MCTP] = "PF_MCTP", 221 }; 222 223 /* 224 * The protocol list. Each protocol is registered in here. 225 */ 226 227 static DEFINE_SPINLOCK(net_family_lock); 228 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly; 229 230 /* 231 * Support routines. 232 * Move socket addresses back and forth across the kernel/user 233 * divide and look after the messy bits. 234 */ 235 236 /** 237 * move_addr_to_kernel - copy a socket address into kernel space 238 * @uaddr: Address in user space 239 * @kaddr: Address in kernel space 240 * @ulen: Length in user space 241 * 242 * The address is copied into kernel space. If the provided address is 243 * too long an error code of -EINVAL is returned. If the copy gives 244 * invalid addresses -EFAULT is returned. On a success 0 is returned. 245 */ 246 247 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr) 248 { 249 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage)) 250 return -EINVAL; 251 if (ulen == 0) 252 return 0; 253 if (copy_from_user(kaddr, uaddr, ulen)) 254 return -EFAULT; 255 return audit_sockaddr(ulen, kaddr); 256 } 257 258 /** 259 * move_addr_to_user - copy an address to user space 260 * @kaddr: kernel space address 261 * @klen: length of address in kernel 262 * @uaddr: user space address 263 * @ulen: pointer to user length field 264 * 265 * The value pointed to by ulen on entry is the buffer length available. 266 * This is overwritten with the buffer space used. -EINVAL is returned 267 * if an overlong buffer is specified or a negative buffer size. -EFAULT 268 * is returned if either the buffer or the length field are not 269 * accessible. 270 * After copying the data up to the limit the user specifies, the true 271 * length of the data is written over the length limit the user 272 * specified. Zero is returned for a success. 273 */ 274 275 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen, 276 void __user *uaddr, int __user *ulen) 277 { 278 int err; 279 int len; 280 281 BUG_ON(klen > sizeof(struct sockaddr_storage)); 282 err = get_user(len, ulen); 283 if (err) 284 return err; 285 if (len > klen) 286 len = klen; 287 if (len < 0) 288 return -EINVAL; 289 if (len) { 290 if (audit_sockaddr(klen, kaddr)) 291 return -ENOMEM; 292 if (copy_to_user(uaddr, kaddr, len)) 293 return -EFAULT; 294 } 295 /* 296 * "fromlen shall refer to the value before truncation.." 297 * 1003.1g 298 */ 299 return __put_user(klen, ulen); 300 } 301 302 static struct kmem_cache *sock_inode_cachep __ro_after_init; 303 304 static struct inode *sock_alloc_inode(struct super_block *sb) 305 { 306 struct socket_alloc *ei; 307 308 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL); 309 if (!ei) 310 return NULL; 311 init_waitqueue_head(&ei->socket.wq.wait); 312 ei->socket.wq.fasync_list = NULL; 313 ei->socket.wq.flags = 0; 314 315 ei->socket.state = SS_UNCONNECTED; 316 ei->socket.flags = 0; 317 ei->socket.ops = NULL; 318 ei->socket.sk = NULL; 319 ei->socket.file = NULL; 320 321 return &ei->vfs_inode; 322 } 323 324 static void sock_free_inode(struct inode *inode) 325 { 326 struct socket_alloc *ei; 327 328 ei = container_of(inode, struct socket_alloc, vfs_inode); 329 kmem_cache_free(sock_inode_cachep, ei); 330 } 331 332 static void init_once(void *foo) 333 { 334 struct socket_alloc *ei = (struct socket_alloc *)foo; 335 336 inode_init_once(&ei->vfs_inode); 337 } 338 339 static void init_inodecache(void) 340 { 341 sock_inode_cachep = kmem_cache_create("sock_inode_cache", 342 sizeof(struct socket_alloc), 343 0, 344 (SLAB_HWCACHE_ALIGN | 345 SLAB_RECLAIM_ACCOUNT | 346 SLAB_ACCOUNT), 347 init_once); 348 BUG_ON(sock_inode_cachep == NULL); 349 } 350 351 static const struct super_operations sockfs_ops = { 352 .alloc_inode = sock_alloc_inode, 353 .free_inode = sock_free_inode, 354 .statfs = simple_statfs, 355 }; 356 357 /* 358 * sockfs_dname() is called from d_path(). 359 */ 360 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen) 361 { 362 return dynamic_dname(buffer, buflen, "socket:[%lu]", 363 d_inode(dentry)->i_ino); 364 } 365 366 static const struct dentry_operations sockfs_dentry_operations = { 367 .d_dname = sockfs_dname, 368 }; 369 370 static int sockfs_xattr_get(const struct xattr_handler *handler, 371 struct dentry *dentry, struct inode *inode, 372 const char *suffix, void *value, size_t size) 373 { 374 if (value) { 375 if (dentry->d_name.len + 1 > size) 376 return -ERANGE; 377 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1); 378 } 379 return dentry->d_name.len + 1; 380 } 381 382 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname" 383 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX) 384 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1) 385 386 static const struct xattr_handler sockfs_xattr_handler = { 387 .name = XATTR_NAME_SOCKPROTONAME, 388 .get = sockfs_xattr_get, 389 }; 390 391 static int sockfs_security_xattr_set(const struct xattr_handler *handler, 392 struct mnt_idmap *idmap, 393 struct dentry *dentry, struct inode *inode, 394 const char *suffix, const void *value, 395 size_t size, int flags) 396 { 397 /* Handled by LSM. */ 398 return -EAGAIN; 399 } 400 401 static const struct xattr_handler sockfs_security_xattr_handler = { 402 .prefix = XATTR_SECURITY_PREFIX, 403 .set = sockfs_security_xattr_set, 404 }; 405 406 static const struct xattr_handler * const sockfs_xattr_handlers[] = { 407 &sockfs_xattr_handler, 408 &sockfs_security_xattr_handler, 409 NULL 410 }; 411 412 static int sockfs_init_fs_context(struct fs_context *fc) 413 { 414 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC); 415 if (!ctx) 416 return -ENOMEM; 417 ctx->ops = &sockfs_ops; 418 ctx->dops = &sockfs_dentry_operations; 419 ctx->xattr = sockfs_xattr_handlers; 420 return 0; 421 } 422 423 static struct vfsmount *sock_mnt __read_mostly; 424 425 static struct file_system_type sock_fs_type = { 426 .name = "sockfs", 427 .init_fs_context = sockfs_init_fs_context, 428 .kill_sb = kill_anon_super, 429 }; 430 431 /* 432 * Obtains the first available file descriptor and sets it up for use. 433 * 434 * These functions create file structures and maps them to fd space 435 * of the current process. On success it returns file descriptor 436 * and file struct implicitly stored in sock->file. 437 * Note that another thread may close file descriptor before we return 438 * from this function. We use the fact that now we do not refer 439 * to socket after mapping. If one day we will need it, this 440 * function will increment ref. count on file by 1. 441 * 442 * In any case returned fd MAY BE not valid! 443 * This race condition is unavoidable 444 * with shared fd spaces, we cannot solve it inside kernel, 445 * but we take care of internal coherence yet. 446 */ 447 448 /** 449 * sock_alloc_file - Bind a &socket to a &file 450 * @sock: socket 451 * @flags: file status flags 452 * @dname: protocol name 453 * 454 * Returns the &file bound with @sock, implicitly storing it 455 * in sock->file. If dname is %NULL, sets to "". 456 * 457 * On failure @sock is released, and an ERR pointer is returned. 458 * 459 * This function uses GFP_KERNEL internally. 460 */ 461 462 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname) 463 { 464 struct file *file; 465 466 if (!dname) 467 dname = sock->sk ? sock->sk->sk_prot_creator->name : ""; 468 469 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname, 470 O_RDWR | (flags & O_NONBLOCK), 471 &socket_file_ops); 472 if (IS_ERR(file)) { 473 sock_release(sock); 474 return file; 475 } 476 477 file->f_mode |= FMODE_NOWAIT; 478 sock->file = file; 479 file->private_data = sock; 480 stream_open(SOCK_INODE(sock), file); 481 return file; 482 } 483 EXPORT_SYMBOL(sock_alloc_file); 484 485 static int sock_map_fd(struct socket *sock, int flags) 486 { 487 struct file *newfile; 488 int fd = get_unused_fd_flags(flags); 489 if (unlikely(fd < 0)) { 490 sock_release(sock); 491 return fd; 492 } 493 494 newfile = sock_alloc_file(sock, flags, NULL); 495 if (!IS_ERR(newfile)) { 496 fd_install(fd, newfile); 497 return fd; 498 } 499 500 put_unused_fd(fd); 501 return PTR_ERR(newfile); 502 } 503 504 /** 505 * sock_from_file - Return the &socket bounded to @file. 506 * @file: file 507 * 508 * On failure returns %NULL. 509 */ 510 511 struct socket *sock_from_file(struct file *file) 512 { 513 if (file->f_op == &socket_file_ops) 514 return file->private_data; /* set in sock_alloc_file */ 515 516 return NULL; 517 } 518 EXPORT_SYMBOL(sock_from_file); 519 520 /** 521 * sockfd_lookup - Go from a file number to its socket slot 522 * @fd: file handle 523 * @err: pointer to an error code return 524 * 525 * The file handle passed in is locked and the socket it is bound 526 * to is returned. If an error occurs the err pointer is overwritten 527 * with a negative errno code and NULL is returned. The function checks 528 * for both invalid handles and passing a handle which is not a socket. 529 * 530 * On a success the socket object pointer is returned. 531 */ 532 533 struct socket *sockfd_lookup(int fd, int *err) 534 { 535 struct file *file; 536 struct socket *sock; 537 538 file = fget(fd); 539 if (!file) { 540 *err = -EBADF; 541 return NULL; 542 } 543 544 sock = sock_from_file(file); 545 if (!sock) { 546 *err = -ENOTSOCK; 547 fput(file); 548 } 549 return sock; 550 } 551 EXPORT_SYMBOL(sockfd_lookup); 552 553 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed) 554 { 555 struct fd f = fdget(fd); 556 struct socket *sock; 557 558 *err = -EBADF; 559 if (f.file) { 560 sock = sock_from_file(f.file); 561 if (likely(sock)) { 562 *fput_needed = f.flags & FDPUT_FPUT; 563 return sock; 564 } 565 *err = -ENOTSOCK; 566 fdput(f); 567 } 568 return NULL; 569 } 570 571 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer, 572 size_t size) 573 { 574 ssize_t len; 575 ssize_t used = 0; 576 577 len = security_inode_listsecurity(d_inode(dentry), buffer, size); 578 if (len < 0) 579 return len; 580 used += len; 581 if (buffer) { 582 if (size < used) 583 return -ERANGE; 584 buffer += len; 585 } 586 587 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1); 588 used += len; 589 if (buffer) { 590 if (size < used) 591 return -ERANGE; 592 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len); 593 buffer += len; 594 } 595 596 return used; 597 } 598 599 static int sockfs_setattr(struct mnt_idmap *idmap, 600 struct dentry *dentry, struct iattr *iattr) 601 { 602 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr); 603 604 if (!err && (iattr->ia_valid & ATTR_UID)) { 605 struct socket *sock = SOCKET_I(d_inode(dentry)); 606 607 if (sock->sk) 608 sock->sk->sk_uid = iattr->ia_uid; 609 else 610 err = -ENOENT; 611 } 612 613 return err; 614 } 615 616 static const struct inode_operations sockfs_inode_ops = { 617 .listxattr = sockfs_listxattr, 618 .setattr = sockfs_setattr, 619 }; 620 621 /** 622 * sock_alloc - allocate a socket 623 * 624 * Allocate a new inode and socket object. The two are bound together 625 * and initialised. The socket is then returned. If we are out of inodes 626 * NULL is returned. This functions uses GFP_KERNEL internally. 627 */ 628 629 struct socket *sock_alloc(void) 630 { 631 struct inode *inode; 632 struct socket *sock; 633 634 inode = new_inode_pseudo(sock_mnt->mnt_sb); 635 if (!inode) 636 return NULL; 637 638 sock = SOCKET_I(inode); 639 640 inode->i_ino = get_next_ino(); 641 inode->i_mode = S_IFSOCK | S_IRWXUGO; 642 inode->i_uid = current_fsuid(); 643 inode->i_gid = current_fsgid(); 644 inode->i_op = &sockfs_inode_ops; 645 646 return sock; 647 } 648 EXPORT_SYMBOL(sock_alloc); 649 650 static void __sock_release(struct socket *sock, struct inode *inode) 651 { 652 const struct proto_ops *ops = READ_ONCE(sock->ops); 653 654 if (ops) { 655 struct module *owner = ops->owner; 656 657 if (inode) 658 inode_lock(inode); 659 ops->release(sock); 660 sock->sk = NULL; 661 if (inode) 662 inode_unlock(inode); 663 sock->ops = NULL; 664 module_put(owner); 665 } 666 667 if (sock->wq.fasync_list) 668 pr_err("%s: fasync list not empty!\n", __func__); 669 670 if (!sock->file) { 671 iput(SOCK_INODE(sock)); 672 return; 673 } 674 sock->file = NULL; 675 } 676 677 /** 678 * sock_release - close a socket 679 * @sock: socket to close 680 * 681 * The socket is released from the protocol stack if it has a release 682 * callback, and the inode is then released if the socket is bound to 683 * an inode not a file. 684 */ 685 void sock_release(struct socket *sock) 686 { 687 __sock_release(sock, NULL); 688 } 689 EXPORT_SYMBOL(sock_release); 690 691 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags) 692 { 693 u8 flags = *tx_flags; 694 695 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) { 696 flags |= SKBTX_HW_TSTAMP; 697 698 /* PTP hardware clocks can provide a free running cycle counter 699 * as a time base for virtual clocks. Tell driver to use the 700 * free running cycle counter for timestamp if socket is bound 701 * to virtual clock. 702 */ 703 if (tsflags & SOF_TIMESTAMPING_BIND_PHC) 704 flags |= SKBTX_HW_TSTAMP_USE_CYCLES; 705 } 706 707 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE) 708 flags |= SKBTX_SW_TSTAMP; 709 710 if (tsflags & SOF_TIMESTAMPING_TX_SCHED) 711 flags |= SKBTX_SCHED_TSTAMP; 712 713 *tx_flags = flags; 714 } 715 EXPORT_SYMBOL(__sock_tx_timestamp); 716 717 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *, 718 size_t)); 719 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *, 720 size_t)); 721 722 static noinline void call_trace_sock_send_length(struct sock *sk, int ret, 723 int flags) 724 { 725 trace_sock_send_length(sk, ret, 0); 726 } 727 728 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg) 729 { 730 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg, 731 inet_sendmsg, sock, msg, 732 msg_data_left(msg)); 733 BUG_ON(ret == -EIOCBQUEUED); 734 735 if (trace_sock_send_length_enabled()) 736 call_trace_sock_send_length(sock->sk, ret, 0); 737 return ret; 738 } 739 740 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg) 741 { 742 int err = security_socket_sendmsg(sock, msg, 743 msg_data_left(msg)); 744 745 return err ?: sock_sendmsg_nosec(sock, msg); 746 } 747 748 /** 749 * sock_sendmsg - send a message through @sock 750 * @sock: socket 751 * @msg: message to send 752 * 753 * Sends @msg through @sock, passing through LSM. 754 * Returns the number of bytes sent, or an error code. 755 */ 756 int sock_sendmsg(struct socket *sock, struct msghdr *msg) 757 { 758 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name; 759 struct sockaddr_storage address; 760 int save_len = msg->msg_namelen; 761 int ret; 762 763 if (msg->msg_name) { 764 memcpy(&address, msg->msg_name, msg->msg_namelen); 765 msg->msg_name = &address; 766 } 767 768 ret = __sock_sendmsg(sock, msg); 769 msg->msg_name = save_addr; 770 msg->msg_namelen = save_len; 771 772 return ret; 773 } 774 EXPORT_SYMBOL(sock_sendmsg); 775 776 /** 777 * kernel_sendmsg - send a message through @sock (kernel-space) 778 * @sock: socket 779 * @msg: message header 780 * @vec: kernel vec 781 * @num: vec array length 782 * @size: total message data size 783 * 784 * Builds the message data with @vec and sends it through @sock. 785 * Returns the number of bytes sent, or an error code. 786 */ 787 788 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, 789 struct kvec *vec, size_t num, size_t size) 790 { 791 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); 792 return sock_sendmsg(sock, msg); 793 } 794 EXPORT_SYMBOL(kernel_sendmsg); 795 796 /** 797 * kernel_sendmsg_locked - send a message through @sock (kernel-space) 798 * @sk: sock 799 * @msg: message header 800 * @vec: output s/g array 801 * @num: output s/g array length 802 * @size: total message data size 803 * 804 * Builds the message data with @vec and sends it through @sock. 805 * Returns the number of bytes sent, or an error code. 806 * Caller must hold @sk. 807 */ 808 809 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg, 810 struct kvec *vec, size_t num, size_t size) 811 { 812 struct socket *sock = sk->sk_socket; 813 const struct proto_ops *ops = READ_ONCE(sock->ops); 814 815 if (!ops->sendmsg_locked) 816 return sock_no_sendmsg_locked(sk, msg, size); 817 818 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size); 819 820 return ops->sendmsg_locked(sk, msg, msg_data_left(msg)); 821 } 822 EXPORT_SYMBOL(kernel_sendmsg_locked); 823 824 static bool skb_is_err_queue(const struct sk_buff *skb) 825 { 826 /* pkt_type of skbs enqueued on the error queue are set to 827 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do 828 * in recvmsg, since skbs received on a local socket will never 829 * have a pkt_type of PACKET_OUTGOING. 830 */ 831 return skb->pkt_type == PACKET_OUTGOING; 832 } 833 834 /* On transmit, software and hardware timestamps are returned independently. 835 * As the two skb clones share the hardware timestamp, which may be updated 836 * before the software timestamp is received, a hardware TX timestamp may be 837 * returned only if there is no software TX timestamp. Ignore false software 838 * timestamps, which may be made in the __sock_recv_timestamp() call when the 839 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a 840 * hardware timestamp. 841 */ 842 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp) 843 { 844 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb); 845 } 846 847 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index) 848 { 849 bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC; 850 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); 851 struct net_device *orig_dev; 852 ktime_t hwtstamp; 853 854 rcu_read_lock(); 855 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 856 if (orig_dev) { 857 *if_index = orig_dev->ifindex; 858 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles); 859 } else { 860 hwtstamp = shhwtstamps->hwtstamp; 861 } 862 rcu_read_unlock(); 863 864 return hwtstamp; 865 } 866 867 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb, 868 int if_index) 869 { 870 struct scm_ts_pktinfo ts_pktinfo; 871 struct net_device *orig_dev; 872 873 if (!skb_mac_header_was_set(skb)) 874 return; 875 876 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo)); 877 878 if (!if_index) { 879 rcu_read_lock(); 880 orig_dev = dev_get_by_napi_id(skb_napi_id(skb)); 881 if (orig_dev) 882 if_index = orig_dev->ifindex; 883 rcu_read_unlock(); 884 } 885 ts_pktinfo.if_index = if_index; 886 887 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb); 888 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO, 889 sizeof(ts_pktinfo), &ts_pktinfo); 890 } 891 892 /* 893 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP) 894 */ 895 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, 896 struct sk_buff *skb) 897 { 898 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP); 899 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW); 900 struct scm_timestamping_internal tss; 901 int empty = 1, false_tstamp = 0; 902 struct skb_shared_hwtstamps *shhwtstamps = 903 skb_hwtstamps(skb); 904 int if_index; 905 ktime_t hwtstamp; 906 u32 tsflags; 907 908 /* Race occurred between timestamp enabling and packet 909 receiving. Fill in the current time for now. */ 910 if (need_software_tstamp && skb->tstamp == 0) { 911 __net_timestamp(skb); 912 false_tstamp = 1; 913 } 914 915 if (need_software_tstamp) { 916 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) { 917 if (new_tstamp) { 918 struct __kernel_sock_timeval tv; 919 920 skb_get_new_timestamp(skb, &tv); 921 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW, 922 sizeof(tv), &tv); 923 } else { 924 struct __kernel_old_timeval tv; 925 926 skb_get_timestamp(skb, &tv); 927 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD, 928 sizeof(tv), &tv); 929 } 930 } else { 931 if (new_tstamp) { 932 struct __kernel_timespec ts; 933 934 skb_get_new_timestampns(skb, &ts); 935 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW, 936 sizeof(ts), &ts); 937 } else { 938 struct __kernel_old_timespec ts; 939 940 skb_get_timestampns(skb, &ts); 941 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD, 942 sizeof(ts), &ts); 943 } 944 } 945 } 946 947 memset(&tss, 0, sizeof(tss)); 948 tsflags = READ_ONCE(sk->sk_tsflags); 949 if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) && 950 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0)) 951 empty = 0; 952 if (shhwtstamps && 953 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) && 954 !skb_is_swtx_tstamp(skb, false_tstamp)) { 955 if_index = 0; 956 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV) 957 hwtstamp = get_timestamp(sk, skb, &if_index); 958 else 959 hwtstamp = shhwtstamps->hwtstamp; 960 961 if (tsflags & SOF_TIMESTAMPING_BIND_PHC) 962 hwtstamp = ptp_convert_timestamp(&hwtstamp, 963 READ_ONCE(sk->sk_bind_phc)); 964 965 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) { 966 empty = 0; 967 968 if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) && 969 !skb_is_err_queue(skb)) 970 put_ts_pktinfo(msg, skb, if_index); 971 } 972 } 973 if (!empty) { 974 if (sock_flag(sk, SOCK_TSTAMP_NEW)) 975 put_cmsg_scm_timestamping64(msg, &tss); 976 else 977 put_cmsg_scm_timestamping(msg, &tss); 978 979 if (skb_is_err_queue(skb) && skb->len && 980 SKB_EXT_ERR(skb)->opt_stats) 981 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS, 982 skb->len, skb->data); 983 } 984 } 985 EXPORT_SYMBOL_GPL(__sock_recv_timestamp); 986 987 #ifdef CONFIG_WIRELESS 988 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, 989 struct sk_buff *skb) 990 { 991 int ack; 992 993 if (!sock_flag(sk, SOCK_WIFI_STATUS)) 994 return; 995 if (!skb->wifi_acked_valid) 996 return; 997 998 ack = skb->wifi_acked; 999 1000 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack); 1001 } 1002 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status); 1003 #endif 1004 1005 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, 1006 struct sk_buff *skb) 1007 { 1008 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount) 1009 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL, 1010 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount); 1011 } 1012 1013 static void sock_recv_mark(struct msghdr *msg, struct sock *sk, 1014 struct sk_buff *skb) 1015 { 1016 if (sock_flag(sk, SOCK_RCVMARK) && skb) { 1017 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */ 1018 __u32 mark = skb->mark; 1019 1020 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark); 1021 } 1022 } 1023 1024 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, 1025 struct sk_buff *skb) 1026 { 1027 sock_recv_timestamp(msg, sk, skb); 1028 sock_recv_drops(msg, sk, skb); 1029 sock_recv_mark(msg, sk, skb); 1030 } 1031 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs); 1032 1033 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *, 1034 size_t, int)); 1035 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *, 1036 size_t, int)); 1037 1038 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags) 1039 { 1040 trace_sock_recv_length(sk, ret, flags); 1041 } 1042 1043 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg, 1044 int flags) 1045 { 1046 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg, 1047 inet6_recvmsg, 1048 inet_recvmsg, sock, msg, 1049 msg_data_left(msg), flags); 1050 if (trace_sock_recv_length_enabled()) 1051 call_trace_sock_recv_length(sock->sk, ret, flags); 1052 return ret; 1053 } 1054 1055 /** 1056 * sock_recvmsg - receive a message from @sock 1057 * @sock: socket 1058 * @msg: message to receive 1059 * @flags: message flags 1060 * 1061 * Receives @msg from @sock, passing through LSM. Returns the total number 1062 * of bytes received, or an error. 1063 */ 1064 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags) 1065 { 1066 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags); 1067 1068 return err ?: sock_recvmsg_nosec(sock, msg, flags); 1069 } 1070 EXPORT_SYMBOL(sock_recvmsg); 1071 1072 /** 1073 * kernel_recvmsg - Receive a message from a socket (kernel space) 1074 * @sock: The socket to receive the message from 1075 * @msg: Received message 1076 * @vec: Input s/g array for message data 1077 * @num: Size of input s/g array 1078 * @size: Number of bytes to read 1079 * @flags: Message flags (MSG_DONTWAIT, etc...) 1080 * 1081 * On return the msg structure contains the scatter/gather array passed in the 1082 * vec argument. The array is modified so that it consists of the unfilled 1083 * portion of the original array. 1084 * 1085 * The returned value is the total number of bytes received, or an error. 1086 */ 1087 1088 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, 1089 struct kvec *vec, size_t num, size_t size, int flags) 1090 { 1091 msg->msg_control_is_user = false; 1092 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size); 1093 return sock_recvmsg(sock, msg, flags); 1094 } 1095 EXPORT_SYMBOL(kernel_recvmsg); 1096 1097 static ssize_t sock_splice_read(struct file *file, loff_t *ppos, 1098 struct pipe_inode_info *pipe, size_t len, 1099 unsigned int flags) 1100 { 1101 struct socket *sock = file->private_data; 1102 const struct proto_ops *ops; 1103 1104 ops = READ_ONCE(sock->ops); 1105 if (unlikely(!ops->splice_read)) 1106 return copy_splice_read(file, ppos, pipe, len, flags); 1107 1108 return ops->splice_read(sock, ppos, pipe, len, flags); 1109 } 1110 1111 static void sock_splice_eof(struct file *file) 1112 { 1113 struct socket *sock = file->private_data; 1114 const struct proto_ops *ops; 1115 1116 ops = READ_ONCE(sock->ops); 1117 if (ops->splice_eof) 1118 ops->splice_eof(sock); 1119 } 1120 1121 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to) 1122 { 1123 struct file *file = iocb->ki_filp; 1124 struct socket *sock = file->private_data; 1125 struct msghdr msg = {.msg_iter = *to, 1126 .msg_iocb = iocb}; 1127 ssize_t res; 1128 1129 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1130 msg.msg_flags = MSG_DONTWAIT; 1131 1132 if (iocb->ki_pos != 0) 1133 return -ESPIPE; 1134 1135 if (!iov_iter_count(to)) /* Match SYS5 behaviour */ 1136 return 0; 1137 1138 res = sock_recvmsg(sock, &msg, msg.msg_flags); 1139 *to = msg.msg_iter; 1140 return res; 1141 } 1142 1143 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from) 1144 { 1145 struct file *file = iocb->ki_filp; 1146 struct socket *sock = file->private_data; 1147 struct msghdr msg = {.msg_iter = *from, 1148 .msg_iocb = iocb}; 1149 ssize_t res; 1150 1151 if (iocb->ki_pos != 0) 1152 return -ESPIPE; 1153 1154 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT)) 1155 msg.msg_flags = MSG_DONTWAIT; 1156 1157 if (sock->type == SOCK_SEQPACKET) 1158 msg.msg_flags |= MSG_EOR; 1159 1160 res = __sock_sendmsg(sock, &msg); 1161 *from = msg.msg_iter; 1162 return res; 1163 } 1164 1165 /* 1166 * Atomic setting of ioctl hooks to avoid race 1167 * with module unload. 1168 */ 1169 1170 static DEFINE_MUTEX(br_ioctl_mutex); 1171 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br, 1172 unsigned int cmd, struct ifreq *ifr, 1173 void __user *uarg); 1174 1175 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, 1176 unsigned int cmd, struct ifreq *ifr, 1177 void __user *uarg)) 1178 { 1179 mutex_lock(&br_ioctl_mutex); 1180 br_ioctl_hook = hook; 1181 mutex_unlock(&br_ioctl_mutex); 1182 } 1183 EXPORT_SYMBOL(brioctl_set); 1184 1185 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, 1186 struct ifreq *ifr, void __user *uarg) 1187 { 1188 int err = -ENOPKG; 1189 1190 if (!br_ioctl_hook) 1191 request_module("bridge"); 1192 1193 mutex_lock(&br_ioctl_mutex); 1194 if (br_ioctl_hook) 1195 err = br_ioctl_hook(net, br, cmd, ifr, uarg); 1196 mutex_unlock(&br_ioctl_mutex); 1197 1198 return err; 1199 } 1200 1201 static DEFINE_MUTEX(vlan_ioctl_mutex); 1202 static int (*vlan_ioctl_hook) (struct net *, void __user *arg); 1203 1204 void vlan_ioctl_set(int (*hook) (struct net *, void __user *)) 1205 { 1206 mutex_lock(&vlan_ioctl_mutex); 1207 vlan_ioctl_hook = hook; 1208 mutex_unlock(&vlan_ioctl_mutex); 1209 } 1210 EXPORT_SYMBOL(vlan_ioctl_set); 1211 1212 static long sock_do_ioctl(struct net *net, struct socket *sock, 1213 unsigned int cmd, unsigned long arg) 1214 { 1215 const struct proto_ops *ops = READ_ONCE(sock->ops); 1216 struct ifreq ifr; 1217 bool need_copyout; 1218 int err; 1219 void __user *argp = (void __user *)arg; 1220 void __user *data; 1221 1222 err = ops->ioctl(sock, cmd, arg); 1223 1224 /* 1225 * If this ioctl is unknown try to hand it down 1226 * to the NIC driver. 1227 */ 1228 if (err != -ENOIOCTLCMD) 1229 return err; 1230 1231 if (!is_socket_ioctl_cmd(cmd)) 1232 return -ENOTTY; 1233 1234 if (get_user_ifreq(&ifr, &data, argp)) 1235 return -EFAULT; 1236 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1237 if (!err && need_copyout) 1238 if (put_user_ifreq(&ifr, argp)) 1239 return -EFAULT; 1240 1241 return err; 1242 } 1243 1244 /* 1245 * With an ioctl, arg may well be a user mode pointer, but we don't know 1246 * what to do with it - that's up to the protocol still. 1247 */ 1248 1249 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg) 1250 { 1251 const struct proto_ops *ops; 1252 struct socket *sock; 1253 struct sock *sk; 1254 void __user *argp = (void __user *)arg; 1255 int pid, err; 1256 struct net *net; 1257 1258 sock = file->private_data; 1259 ops = READ_ONCE(sock->ops); 1260 sk = sock->sk; 1261 net = sock_net(sk); 1262 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) { 1263 struct ifreq ifr; 1264 void __user *data; 1265 bool need_copyout; 1266 if (get_user_ifreq(&ifr, &data, argp)) 1267 return -EFAULT; 1268 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout); 1269 if (!err && need_copyout) 1270 if (put_user_ifreq(&ifr, argp)) 1271 return -EFAULT; 1272 } else 1273 #ifdef CONFIG_WEXT_CORE 1274 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) { 1275 err = wext_handle_ioctl(net, cmd, argp); 1276 } else 1277 #endif 1278 switch (cmd) { 1279 case FIOSETOWN: 1280 case SIOCSPGRP: 1281 err = -EFAULT; 1282 if (get_user(pid, (int __user *)argp)) 1283 break; 1284 err = f_setown(sock->file, pid, 1); 1285 break; 1286 case FIOGETOWN: 1287 case SIOCGPGRP: 1288 err = put_user(f_getown(sock->file), 1289 (int __user *)argp); 1290 break; 1291 case SIOCGIFBR: 1292 case SIOCSIFBR: 1293 case SIOCBRADDBR: 1294 case SIOCBRDELBR: 1295 err = br_ioctl_call(net, NULL, cmd, NULL, argp); 1296 break; 1297 case SIOCGIFVLAN: 1298 case SIOCSIFVLAN: 1299 err = -ENOPKG; 1300 if (!vlan_ioctl_hook) 1301 request_module("8021q"); 1302 1303 mutex_lock(&vlan_ioctl_mutex); 1304 if (vlan_ioctl_hook) 1305 err = vlan_ioctl_hook(net, argp); 1306 mutex_unlock(&vlan_ioctl_mutex); 1307 break; 1308 case SIOCGSKNS: 1309 err = -EPERM; 1310 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1311 break; 1312 1313 err = open_related_ns(&net->ns, get_net_ns); 1314 break; 1315 case SIOCGSTAMP_OLD: 1316 case SIOCGSTAMPNS_OLD: 1317 if (!ops->gettstamp) { 1318 err = -ENOIOCTLCMD; 1319 break; 1320 } 1321 err = ops->gettstamp(sock, argp, 1322 cmd == SIOCGSTAMP_OLD, 1323 !IS_ENABLED(CONFIG_64BIT)); 1324 break; 1325 case SIOCGSTAMP_NEW: 1326 case SIOCGSTAMPNS_NEW: 1327 if (!ops->gettstamp) { 1328 err = -ENOIOCTLCMD; 1329 break; 1330 } 1331 err = ops->gettstamp(sock, argp, 1332 cmd == SIOCGSTAMP_NEW, 1333 false); 1334 break; 1335 1336 case SIOCGIFCONF: 1337 err = dev_ifconf(net, argp); 1338 break; 1339 1340 default: 1341 err = sock_do_ioctl(net, sock, cmd, arg); 1342 break; 1343 } 1344 return err; 1345 } 1346 1347 /** 1348 * sock_create_lite - creates a socket 1349 * @family: protocol family (AF_INET, ...) 1350 * @type: communication type (SOCK_STREAM, ...) 1351 * @protocol: protocol (0, ...) 1352 * @res: new socket 1353 * 1354 * Creates a new socket and assigns it to @res, passing through LSM. 1355 * The new socket initialization is not complete, see kernel_accept(). 1356 * Returns 0 or an error. On failure @res is set to %NULL. 1357 * This function internally uses GFP_KERNEL. 1358 */ 1359 1360 int sock_create_lite(int family, int type, int protocol, struct socket **res) 1361 { 1362 int err; 1363 struct socket *sock = NULL; 1364 1365 err = security_socket_create(family, type, protocol, 1); 1366 if (err) 1367 goto out; 1368 1369 sock = sock_alloc(); 1370 if (!sock) { 1371 err = -ENOMEM; 1372 goto out; 1373 } 1374 1375 sock->type = type; 1376 err = security_socket_post_create(sock, family, type, protocol, 1); 1377 if (err) 1378 goto out_release; 1379 1380 out: 1381 *res = sock; 1382 return err; 1383 out_release: 1384 sock_release(sock); 1385 sock = NULL; 1386 goto out; 1387 } 1388 EXPORT_SYMBOL(sock_create_lite); 1389 1390 /* No kernel lock held - perfect */ 1391 static __poll_t sock_poll(struct file *file, poll_table *wait) 1392 { 1393 struct socket *sock = file->private_data; 1394 const struct proto_ops *ops = READ_ONCE(sock->ops); 1395 __poll_t events = poll_requested_events(wait), flag = 0; 1396 1397 if (!ops->poll) 1398 return 0; 1399 1400 if (sk_can_busy_loop(sock->sk)) { 1401 /* poll once if requested by the syscall */ 1402 if (events & POLL_BUSY_LOOP) 1403 sk_busy_loop(sock->sk, 1); 1404 1405 /* if this socket can poll_ll, tell the system call */ 1406 flag = POLL_BUSY_LOOP; 1407 } 1408 1409 return ops->poll(file, sock, wait) | flag; 1410 } 1411 1412 static int sock_mmap(struct file *file, struct vm_area_struct *vma) 1413 { 1414 struct socket *sock = file->private_data; 1415 1416 return READ_ONCE(sock->ops)->mmap(file, sock, vma); 1417 } 1418 1419 static int sock_close(struct inode *inode, struct file *filp) 1420 { 1421 __sock_release(SOCKET_I(inode), inode); 1422 return 0; 1423 } 1424 1425 /* 1426 * Update the socket async list 1427 * 1428 * Fasync_list locking strategy. 1429 * 1430 * 1. fasync_list is modified only under process context socket lock 1431 * i.e. under semaphore. 1432 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock) 1433 * or under socket lock 1434 */ 1435 1436 static int sock_fasync(int fd, struct file *filp, int on) 1437 { 1438 struct socket *sock = filp->private_data; 1439 struct sock *sk = sock->sk; 1440 struct socket_wq *wq = &sock->wq; 1441 1442 if (sk == NULL) 1443 return -EINVAL; 1444 1445 lock_sock(sk); 1446 fasync_helper(fd, filp, on, &wq->fasync_list); 1447 1448 if (!wq->fasync_list) 1449 sock_reset_flag(sk, SOCK_FASYNC); 1450 else 1451 sock_set_flag(sk, SOCK_FASYNC); 1452 1453 release_sock(sk); 1454 return 0; 1455 } 1456 1457 /* This function may be called only under rcu_lock */ 1458 1459 int sock_wake_async(struct socket_wq *wq, int how, int band) 1460 { 1461 if (!wq || !wq->fasync_list) 1462 return -1; 1463 1464 switch (how) { 1465 case SOCK_WAKE_WAITD: 1466 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags)) 1467 break; 1468 goto call_kill; 1469 case SOCK_WAKE_SPACE: 1470 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags)) 1471 break; 1472 fallthrough; 1473 case SOCK_WAKE_IO: 1474 call_kill: 1475 kill_fasync(&wq->fasync_list, SIGIO, band); 1476 break; 1477 case SOCK_WAKE_URG: 1478 kill_fasync(&wq->fasync_list, SIGURG, band); 1479 } 1480 1481 return 0; 1482 } 1483 EXPORT_SYMBOL(sock_wake_async); 1484 1485 /** 1486 * __sock_create - creates a socket 1487 * @net: net namespace 1488 * @family: protocol family (AF_INET, ...) 1489 * @type: communication type (SOCK_STREAM, ...) 1490 * @protocol: protocol (0, ...) 1491 * @res: new socket 1492 * @kern: boolean for kernel space sockets 1493 * 1494 * Creates a new socket and assigns it to @res, passing through LSM. 1495 * Returns 0 or an error. On failure @res is set to %NULL. @kern must 1496 * be set to true if the socket resides in kernel space. 1497 * This function internally uses GFP_KERNEL. 1498 */ 1499 1500 int __sock_create(struct net *net, int family, int type, int protocol, 1501 struct socket **res, int kern) 1502 { 1503 int err; 1504 struct socket *sock; 1505 const struct net_proto_family *pf; 1506 1507 /* 1508 * Check protocol is in range 1509 */ 1510 if (family < 0 || family >= NPROTO) 1511 return -EAFNOSUPPORT; 1512 if (type < 0 || type >= SOCK_MAX) 1513 return -EINVAL; 1514 1515 /* Compatibility. 1516 1517 This uglymoron is moved from INET layer to here to avoid 1518 deadlock in module load. 1519 */ 1520 if (family == PF_INET && type == SOCK_PACKET) { 1521 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n", 1522 current->comm); 1523 family = PF_PACKET; 1524 } 1525 1526 err = security_socket_create(family, type, protocol, kern); 1527 if (err) 1528 return err; 1529 1530 /* 1531 * Allocate the socket and allow the family to set things up. if 1532 * the protocol is 0, the family is instructed to select an appropriate 1533 * default. 1534 */ 1535 sock = sock_alloc(); 1536 if (!sock) { 1537 net_warn_ratelimited("socket: no more sockets\n"); 1538 return -ENFILE; /* Not exactly a match, but its the 1539 closest posix thing */ 1540 } 1541 1542 sock->type = type; 1543 1544 #ifdef CONFIG_MODULES 1545 /* Attempt to load a protocol module if the find failed. 1546 * 1547 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user 1548 * requested real, full-featured networking support upon configuration. 1549 * Otherwise module support will break! 1550 */ 1551 if (rcu_access_pointer(net_families[family]) == NULL) 1552 request_module("net-pf-%d", family); 1553 #endif 1554 1555 rcu_read_lock(); 1556 pf = rcu_dereference(net_families[family]); 1557 err = -EAFNOSUPPORT; 1558 if (!pf) 1559 goto out_release; 1560 1561 /* 1562 * We will call the ->create function, that possibly is in a loadable 1563 * module, so we have to bump that loadable module refcnt first. 1564 */ 1565 if (!try_module_get(pf->owner)) 1566 goto out_release; 1567 1568 /* Now protected by module ref count */ 1569 rcu_read_unlock(); 1570 1571 err = pf->create(net, sock, protocol, kern); 1572 if (err < 0) 1573 goto out_module_put; 1574 1575 /* 1576 * Now to bump the refcnt of the [loadable] module that owns this 1577 * socket at sock_release time we decrement its refcnt. 1578 */ 1579 if (!try_module_get(sock->ops->owner)) 1580 goto out_module_busy; 1581 1582 /* 1583 * Now that we're done with the ->create function, the [loadable] 1584 * module can have its refcnt decremented 1585 */ 1586 module_put(pf->owner); 1587 err = security_socket_post_create(sock, family, type, protocol, kern); 1588 if (err) 1589 goto out_sock_release; 1590 *res = sock; 1591 1592 return 0; 1593 1594 out_module_busy: 1595 err = -EAFNOSUPPORT; 1596 out_module_put: 1597 sock->ops = NULL; 1598 module_put(pf->owner); 1599 out_sock_release: 1600 sock_release(sock); 1601 return err; 1602 1603 out_release: 1604 rcu_read_unlock(); 1605 goto out_sock_release; 1606 } 1607 EXPORT_SYMBOL(__sock_create); 1608 1609 /** 1610 * sock_create - creates a socket 1611 * @family: protocol family (AF_INET, ...) 1612 * @type: communication type (SOCK_STREAM, ...) 1613 * @protocol: protocol (0, ...) 1614 * @res: new socket 1615 * 1616 * A wrapper around __sock_create(). 1617 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1618 */ 1619 1620 int sock_create(int family, int type, int protocol, struct socket **res) 1621 { 1622 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0); 1623 } 1624 EXPORT_SYMBOL(sock_create); 1625 1626 /** 1627 * sock_create_kern - creates a socket (kernel space) 1628 * @net: net namespace 1629 * @family: protocol family (AF_INET, ...) 1630 * @type: communication type (SOCK_STREAM, ...) 1631 * @protocol: protocol (0, ...) 1632 * @res: new socket 1633 * 1634 * A wrapper around __sock_create(). 1635 * Returns 0 or an error. This function internally uses GFP_KERNEL. 1636 */ 1637 1638 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res) 1639 { 1640 return __sock_create(net, family, type, protocol, res, 1); 1641 } 1642 EXPORT_SYMBOL(sock_create_kern); 1643 1644 static struct socket *__sys_socket_create(int family, int type, int protocol) 1645 { 1646 struct socket *sock; 1647 int retval; 1648 1649 /* Check the SOCK_* constants for consistency. */ 1650 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC); 1651 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK); 1652 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK); 1653 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK); 1654 1655 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1656 return ERR_PTR(-EINVAL); 1657 type &= SOCK_TYPE_MASK; 1658 1659 retval = sock_create(family, type, protocol, &sock); 1660 if (retval < 0) 1661 return ERR_PTR(retval); 1662 1663 return sock; 1664 } 1665 1666 struct file *__sys_socket_file(int family, int type, int protocol) 1667 { 1668 struct socket *sock; 1669 int flags; 1670 1671 sock = __sys_socket_create(family, type, protocol); 1672 if (IS_ERR(sock)) 1673 return ERR_CAST(sock); 1674 1675 flags = type & ~SOCK_TYPE_MASK; 1676 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1677 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1678 1679 return sock_alloc_file(sock, flags, NULL); 1680 } 1681 1682 /* A hook for bpf progs to attach to and update socket protocol. 1683 * 1684 * A static noinline declaration here could cause the compiler to 1685 * optimize away the function. A global noinline declaration will 1686 * keep the definition, but may optimize away the callsite. 1687 * Therefore, __weak is needed to ensure that the call is still 1688 * emitted, by telling the compiler that we don't know what the 1689 * function might eventually be. 1690 */ 1691 1692 __bpf_hook_start(); 1693 1694 __weak noinline int update_socket_protocol(int family, int type, int protocol) 1695 { 1696 return protocol; 1697 } 1698 1699 __bpf_hook_end(); 1700 1701 int __sys_socket(int family, int type, int protocol) 1702 { 1703 struct socket *sock; 1704 int flags; 1705 1706 sock = __sys_socket_create(family, type, 1707 update_socket_protocol(family, type, protocol)); 1708 if (IS_ERR(sock)) 1709 return PTR_ERR(sock); 1710 1711 flags = type & ~SOCK_TYPE_MASK; 1712 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1713 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1714 1715 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK)); 1716 } 1717 1718 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol) 1719 { 1720 return __sys_socket(family, type, protocol); 1721 } 1722 1723 /* 1724 * Create a pair of connected sockets. 1725 */ 1726 1727 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec) 1728 { 1729 struct socket *sock1, *sock2; 1730 int fd1, fd2, err; 1731 struct file *newfile1, *newfile2; 1732 int flags; 1733 1734 flags = type & ~SOCK_TYPE_MASK; 1735 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1736 return -EINVAL; 1737 type &= SOCK_TYPE_MASK; 1738 1739 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1740 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1741 1742 /* 1743 * reserve descriptors and make sure we won't fail 1744 * to return them to userland. 1745 */ 1746 fd1 = get_unused_fd_flags(flags); 1747 if (unlikely(fd1 < 0)) 1748 return fd1; 1749 1750 fd2 = get_unused_fd_flags(flags); 1751 if (unlikely(fd2 < 0)) { 1752 put_unused_fd(fd1); 1753 return fd2; 1754 } 1755 1756 err = put_user(fd1, &usockvec[0]); 1757 if (err) 1758 goto out; 1759 1760 err = put_user(fd2, &usockvec[1]); 1761 if (err) 1762 goto out; 1763 1764 /* 1765 * Obtain the first socket and check if the underlying protocol 1766 * supports the socketpair call. 1767 */ 1768 1769 err = sock_create(family, type, protocol, &sock1); 1770 if (unlikely(err < 0)) 1771 goto out; 1772 1773 err = sock_create(family, type, protocol, &sock2); 1774 if (unlikely(err < 0)) { 1775 sock_release(sock1); 1776 goto out; 1777 } 1778 1779 err = security_socket_socketpair(sock1, sock2); 1780 if (unlikely(err)) { 1781 sock_release(sock2); 1782 sock_release(sock1); 1783 goto out; 1784 } 1785 1786 err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2); 1787 if (unlikely(err < 0)) { 1788 sock_release(sock2); 1789 sock_release(sock1); 1790 goto out; 1791 } 1792 1793 newfile1 = sock_alloc_file(sock1, flags, NULL); 1794 if (IS_ERR(newfile1)) { 1795 err = PTR_ERR(newfile1); 1796 sock_release(sock2); 1797 goto out; 1798 } 1799 1800 newfile2 = sock_alloc_file(sock2, flags, NULL); 1801 if (IS_ERR(newfile2)) { 1802 err = PTR_ERR(newfile2); 1803 fput(newfile1); 1804 goto out; 1805 } 1806 1807 audit_fd_pair(fd1, fd2); 1808 1809 fd_install(fd1, newfile1); 1810 fd_install(fd2, newfile2); 1811 return 0; 1812 1813 out: 1814 put_unused_fd(fd2); 1815 put_unused_fd(fd1); 1816 return err; 1817 } 1818 1819 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, 1820 int __user *, usockvec) 1821 { 1822 return __sys_socketpair(family, type, protocol, usockvec); 1823 } 1824 1825 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address, 1826 int addrlen) 1827 { 1828 int err; 1829 1830 err = security_socket_bind(sock, (struct sockaddr *)address, 1831 addrlen); 1832 if (!err) 1833 err = READ_ONCE(sock->ops)->bind(sock, 1834 (struct sockaddr *)address, 1835 addrlen); 1836 return err; 1837 } 1838 1839 /* 1840 * Bind a name to a socket. Nothing much to do here since it's 1841 * the protocol's responsibility to handle the local address. 1842 * 1843 * We move the socket address to kernel space before we call 1844 * the protocol layer (having also checked the address is ok). 1845 */ 1846 1847 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) 1848 { 1849 struct socket *sock; 1850 struct sockaddr_storage address; 1851 int err, fput_needed; 1852 1853 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1854 if (sock) { 1855 err = move_addr_to_kernel(umyaddr, addrlen, &address); 1856 if (!err) 1857 err = __sys_bind_socket(sock, &address, addrlen); 1858 fput_light(sock->file, fput_needed); 1859 } 1860 return err; 1861 } 1862 1863 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) 1864 { 1865 return __sys_bind(fd, umyaddr, addrlen); 1866 } 1867 1868 /* 1869 * Perform a listen. Basically, we allow the protocol to do anything 1870 * necessary for a listen, and if that works, we mark the socket as 1871 * ready for listening. 1872 */ 1873 int __sys_listen_socket(struct socket *sock, int backlog) 1874 { 1875 int somaxconn, err; 1876 1877 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); 1878 if ((unsigned int)backlog > somaxconn) 1879 backlog = somaxconn; 1880 1881 err = security_socket_listen(sock, backlog); 1882 if (!err) 1883 err = READ_ONCE(sock->ops)->listen(sock, backlog); 1884 return err; 1885 } 1886 1887 int __sys_listen(int fd, int backlog) 1888 { 1889 struct socket *sock; 1890 int err, fput_needed; 1891 1892 sock = sockfd_lookup_light(fd, &err, &fput_needed); 1893 if (sock) { 1894 err = __sys_listen_socket(sock, backlog); 1895 fput_light(sock->file, fput_needed); 1896 } 1897 return err; 1898 } 1899 1900 SYSCALL_DEFINE2(listen, int, fd, int, backlog) 1901 { 1902 return __sys_listen(fd, backlog); 1903 } 1904 1905 struct file *do_accept(struct file *file, struct proto_accept_arg *arg, 1906 struct sockaddr __user *upeer_sockaddr, 1907 int __user *upeer_addrlen, int flags) 1908 { 1909 struct socket *sock, *newsock; 1910 struct file *newfile; 1911 int err, len; 1912 struct sockaddr_storage address; 1913 const struct proto_ops *ops; 1914 1915 sock = sock_from_file(file); 1916 if (!sock) 1917 return ERR_PTR(-ENOTSOCK); 1918 1919 newsock = sock_alloc(); 1920 if (!newsock) 1921 return ERR_PTR(-ENFILE); 1922 ops = READ_ONCE(sock->ops); 1923 1924 newsock->type = sock->type; 1925 newsock->ops = ops; 1926 1927 /* 1928 * We don't need try_module_get here, as the listening socket (sock) 1929 * has the protocol module (sock->ops->owner) held. 1930 */ 1931 __module_get(ops->owner); 1932 1933 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name); 1934 if (IS_ERR(newfile)) 1935 return newfile; 1936 1937 err = security_socket_accept(sock, newsock); 1938 if (err) 1939 goto out_fd; 1940 1941 arg->flags |= sock->file->f_flags; 1942 err = ops->accept(sock, newsock, arg); 1943 if (err < 0) 1944 goto out_fd; 1945 1946 if (upeer_sockaddr) { 1947 len = ops->getname(newsock, (struct sockaddr *)&address, 2); 1948 if (len < 0) { 1949 err = -ECONNABORTED; 1950 goto out_fd; 1951 } 1952 err = move_addr_to_user(&address, 1953 len, upeer_sockaddr, upeer_addrlen); 1954 if (err < 0) 1955 goto out_fd; 1956 } 1957 1958 /* File flags are not inherited via accept() unlike another OSes. */ 1959 return newfile; 1960 out_fd: 1961 fput(newfile); 1962 return ERR_PTR(err); 1963 } 1964 1965 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr, 1966 int __user *upeer_addrlen, int flags) 1967 { 1968 struct proto_accept_arg arg = { }; 1969 struct file *newfile; 1970 int newfd; 1971 1972 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK)) 1973 return -EINVAL; 1974 1975 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK)) 1976 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK; 1977 1978 newfd = get_unused_fd_flags(flags); 1979 if (unlikely(newfd < 0)) 1980 return newfd; 1981 1982 newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen, 1983 flags); 1984 if (IS_ERR(newfile)) { 1985 put_unused_fd(newfd); 1986 return PTR_ERR(newfile); 1987 } 1988 fd_install(newfd, newfile); 1989 return newfd; 1990 } 1991 1992 /* 1993 * For accept, we attempt to create a new socket, set up the link 1994 * with the client, wake up the client, then return the new 1995 * connected fd. We collect the address of the connector in kernel 1996 * space and move it to user at the very end. This is unclean because 1997 * we open the socket then return an error. 1998 * 1999 * 1003.1g adds the ability to recvmsg() to query connection pending 2000 * status to recvmsg. We need to add that support in a way thats 2001 * clean when we restructure accept also. 2002 */ 2003 2004 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, 2005 int __user *upeer_addrlen, int flags) 2006 { 2007 int ret = -EBADF; 2008 struct fd f; 2009 2010 f = fdget(fd); 2011 if (f.file) { 2012 ret = __sys_accept4_file(f.file, upeer_sockaddr, 2013 upeer_addrlen, flags); 2014 fdput(f); 2015 } 2016 2017 return ret; 2018 } 2019 2020 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr, 2021 int __user *, upeer_addrlen, int, flags) 2022 { 2023 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags); 2024 } 2025 2026 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr, 2027 int __user *, upeer_addrlen) 2028 { 2029 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0); 2030 } 2031 2032 /* 2033 * Attempt to connect to a socket with the server address. The address 2034 * is in user space so we verify it is OK and move it to kernel space. 2035 * 2036 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to 2037 * break bindings 2038 * 2039 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and 2040 * other SEQPACKET protocols that take time to connect() as it doesn't 2041 * include the -EINPROGRESS status for such sockets. 2042 */ 2043 2044 int __sys_connect_file(struct file *file, struct sockaddr_storage *address, 2045 int addrlen, int file_flags) 2046 { 2047 struct socket *sock; 2048 int err; 2049 2050 sock = sock_from_file(file); 2051 if (!sock) { 2052 err = -ENOTSOCK; 2053 goto out; 2054 } 2055 2056 err = 2057 security_socket_connect(sock, (struct sockaddr *)address, addrlen); 2058 if (err) 2059 goto out; 2060 2061 err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address, 2062 addrlen, sock->file->f_flags | file_flags); 2063 out: 2064 return err; 2065 } 2066 2067 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen) 2068 { 2069 int ret = -EBADF; 2070 struct fd f; 2071 2072 f = fdget(fd); 2073 if (f.file) { 2074 struct sockaddr_storage address; 2075 2076 ret = move_addr_to_kernel(uservaddr, addrlen, &address); 2077 if (!ret) 2078 ret = __sys_connect_file(f.file, &address, addrlen, 0); 2079 fdput(f); 2080 } 2081 2082 return ret; 2083 } 2084 2085 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr, 2086 int, addrlen) 2087 { 2088 return __sys_connect(fd, uservaddr, addrlen); 2089 } 2090 2091 /* 2092 * Get the local address ('name') of a socket object. Move the obtained 2093 * name to user space. 2094 */ 2095 2096 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, 2097 int __user *usockaddr_len) 2098 { 2099 struct socket *sock; 2100 struct sockaddr_storage address; 2101 int err, fput_needed; 2102 2103 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2104 if (!sock) 2105 goto out; 2106 2107 err = security_socket_getsockname(sock); 2108 if (err) 2109 goto out_put; 2110 2111 err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0); 2112 if (err < 0) 2113 goto out_put; 2114 /* "err" is actually length in this case */ 2115 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len); 2116 2117 out_put: 2118 fput_light(sock->file, fput_needed); 2119 out: 2120 return err; 2121 } 2122 2123 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr, 2124 int __user *, usockaddr_len) 2125 { 2126 return __sys_getsockname(fd, usockaddr, usockaddr_len); 2127 } 2128 2129 /* 2130 * Get the remote address ('name') of a socket object. Move the obtained 2131 * name to user space. 2132 */ 2133 2134 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, 2135 int __user *usockaddr_len) 2136 { 2137 struct socket *sock; 2138 struct sockaddr_storage address; 2139 int err, fput_needed; 2140 2141 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2142 if (sock != NULL) { 2143 const struct proto_ops *ops = READ_ONCE(sock->ops); 2144 2145 err = security_socket_getpeername(sock); 2146 if (err) { 2147 fput_light(sock->file, fput_needed); 2148 return err; 2149 } 2150 2151 err = ops->getname(sock, (struct sockaddr *)&address, 1); 2152 if (err >= 0) 2153 /* "err" is actually length in this case */ 2154 err = move_addr_to_user(&address, err, usockaddr, 2155 usockaddr_len); 2156 fput_light(sock->file, fput_needed); 2157 } 2158 return err; 2159 } 2160 2161 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr, 2162 int __user *, usockaddr_len) 2163 { 2164 return __sys_getpeername(fd, usockaddr, usockaddr_len); 2165 } 2166 2167 /* 2168 * Send a datagram to a given address. We move the address into kernel 2169 * space and check the user space data area is readable before invoking 2170 * the protocol. 2171 */ 2172 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags, 2173 struct sockaddr __user *addr, int addr_len) 2174 { 2175 struct socket *sock; 2176 struct sockaddr_storage address; 2177 int err; 2178 struct msghdr msg; 2179 int fput_needed; 2180 2181 err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter); 2182 if (unlikely(err)) 2183 return err; 2184 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2185 if (!sock) 2186 goto out; 2187 2188 msg.msg_name = NULL; 2189 msg.msg_control = NULL; 2190 msg.msg_controllen = 0; 2191 msg.msg_namelen = 0; 2192 msg.msg_ubuf = NULL; 2193 if (addr) { 2194 err = move_addr_to_kernel(addr, addr_len, &address); 2195 if (err < 0) 2196 goto out_put; 2197 msg.msg_name = (struct sockaddr *)&address; 2198 msg.msg_namelen = addr_len; 2199 } 2200 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; 2201 if (sock->file->f_flags & O_NONBLOCK) 2202 flags |= MSG_DONTWAIT; 2203 msg.msg_flags = flags; 2204 err = __sock_sendmsg(sock, &msg); 2205 2206 out_put: 2207 fput_light(sock->file, fput_needed); 2208 out: 2209 return err; 2210 } 2211 2212 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len, 2213 unsigned int, flags, struct sockaddr __user *, addr, 2214 int, addr_len) 2215 { 2216 return __sys_sendto(fd, buff, len, flags, addr, addr_len); 2217 } 2218 2219 /* 2220 * Send a datagram down a socket. 2221 */ 2222 2223 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, 2224 unsigned int, flags) 2225 { 2226 return __sys_sendto(fd, buff, len, flags, NULL, 0); 2227 } 2228 2229 /* 2230 * Receive a frame from the socket and optionally record the address of the 2231 * sender. We verify the buffers are writable and if needed move the 2232 * sender address from kernel to user space. 2233 */ 2234 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, 2235 struct sockaddr __user *addr, int __user *addr_len) 2236 { 2237 struct sockaddr_storage address; 2238 struct msghdr msg = { 2239 /* Save some cycles and don't copy the address if not needed */ 2240 .msg_name = addr ? (struct sockaddr *)&address : NULL, 2241 }; 2242 struct socket *sock; 2243 int err, err2; 2244 int fput_needed; 2245 2246 err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter); 2247 if (unlikely(err)) 2248 return err; 2249 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2250 if (!sock) 2251 goto out; 2252 2253 if (sock->file->f_flags & O_NONBLOCK) 2254 flags |= MSG_DONTWAIT; 2255 err = sock_recvmsg(sock, &msg, flags); 2256 2257 if (err >= 0 && addr != NULL) { 2258 err2 = move_addr_to_user(&address, 2259 msg.msg_namelen, addr, addr_len); 2260 if (err2 < 0) 2261 err = err2; 2262 } 2263 2264 fput_light(sock->file, fput_needed); 2265 out: 2266 return err; 2267 } 2268 2269 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size, 2270 unsigned int, flags, struct sockaddr __user *, addr, 2271 int __user *, addr_len) 2272 { 2273 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len); 2274 } 2275 2276 /* 2277 * Receive a datagram from a socket. 2278 */ 2279 2280 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size, 2281 unsigned int, flags) 2282 { 2283 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL); 2284 } 2285 2286 static bool sock_use_custom_sol_socket(const struct socket *sock) 2287 { 2288 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags); 2289 } 2290 2291 int do_sock_setsockopt(struct socket *sock, bool compat, int level, 2292 int optname, sockptr_t optval, int optlen) 2293 { 2294 const struct proto_ops *ops; 2295 char *kernel_optval = NULL; 2296 int err; 2297 2298 if (optlen < 0) 2299 return -EINVAL; 2300 2301 err = security_socket_setsockopt(sock, level, optname); 2302 if (err) 2303 goto out_put; 2304 2305 if (!compat) 2306 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname, 2307 optval, &optlen, 2308 &kernel_optval); 2309 if (err < 0) 2310 goto out_put; 2311 if (err > 0) { 2312 err = 0; 2313 goto out_put; 2314 } 2315 2316 if (kernel_optval) 2317 optval = KERNEL_SOCKPTR(kernel_optval); 2318 ops = READ_ONCE(sock->ops); 2319 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock)) 2320 err = sock_setsockopt(sock, level, optname, optval, optlen); 2321 else if (unlikely(!ops->setsockopt)) 2322 err = -EOPNOTSUPP; 2323 else 2324 err = ops->setsockopt(sock, level, optname, optval, 2325 optlen); 2326 kfree(kernel_optval); 2327 out_put: 2328 return err; 2329 } 2330 EXPORT_SYMBOL(do_sock_setsockopt); 2331 2332 /* Set a socket option. Because we don't know the option lengths we have 2333 * to pass the user mode parameter for the protocols to sort out. 2334 */ 2335 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval, 2336 int optlen) 2337 { 2338 sockptr_t optval = USER_SOCKPTR(user_optval); 2339 bool compat = in_compat_syscall(); 2340 int err, fput_needed; 2341 struct socket *sock; 2342 2343 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2344 if (!sock) 2345 return err; 2346 2347 err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen); 2348 2349 fput_light(sock->file, fput_needed); 2350 return err; 2351 } 2352 2353 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname, 2354 char __user *, optval, int, optlen) 2355 { 2356 return __sys_setsockopt(fd, level, optname, optval, optlen); 2357 } 2358 2359 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level, 2360 int optname)); 2361 2362 int do_sock_getsockopt(struct socket *sock, bool compat, int level, 2363 int optname, sockptr_t optval, sockptr_t optlen) 2364 { 2365 int max_optlen __maybe_unused = 0; 2366 const struct proto_ops *ops; 2367 int err; 2368 2369 err = security_socket_getsockopt(sock, level, optname); 2370 if (err) 2371 return err; 2372 2373 if (!compat) 2374 copy_from_sockptr(&max_optlen, optlen, sizeof(int)); 2375 2376 ops = READ_ONCE(sock->ops); 2377 if (level == SOL_SOCKET) { 2378 err = sk_getsockopt(sock->sk, level, optname, optval, optlen); 2379 } else if (unlikely(!ops->getsockopt)) { 2380 err = -EOPNOTSUPP; 2381 } else { 2382 if (WARN_ONCE(optval.is_kernel || optlen.is_kernel, 2383 "Invalid argument type")) 2384 return -EOPNOTSUPP; 2385 2386 err = ops->getsockopt(sock, level, optname, optval.user, 2387 optlen.user); 2388 } 2389 2390 if (!compat) 2391 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname, 2392 optval, optlen, max_optlen, 2393 err); 2394 2395 return err; 2396 } 2397 EXPORT_SYMBOL(do_sock_getsockopt); 2398 2399 /* 2400 * Get a socket option. Because we don't know the option lengths we have 2401 * to pass a user mode parameter for the protocols to sort out. 2402 */ 2403 int __sys_getsockopt(int fd, int level, int optname, char __user *optval, 2404 int __user *optlen) 2405 { 2406 int err, fput_needed; 2407 struct socket *sock; 2408 bool compat; 2409 2410 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2411 if (!sock) 2412 return err; 2413 2414 compat = in_compat_syscall(); 2415 err = do_sock_getsockopt(sock, compat, level, optname, 2416 USER_SOCKPTR(optval), USER_SOCKPTR(optlen)); 2417 2418 fput_light(sock->file, fput_needed); 2419 return err; 2420 } 2421 2422 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname, 2423 char __user *, optval, int __user *, optlen) 2424 { 2425 return __sys_getsockopt(fd, level, optname, optval, optlen); 2426 } 2427 2428 /* 2429 * Shutdown a socket. 2430 */ 2431 2432 int __sys_shutdown_sock(struct socket *sock, int how) 2433 { 2434 int err; 2435 2436 err = security_socket_shutdown(sock, how); 2437 if (!err) 2438 err = READ_ONCE(sock->ops)->shutdown(sock, how); 2439 2440 return err; 2441 } 2442 2443 int __sys_shutdown(int fd, int how) 2444 { 2445 int err, fput_needed; 2446 struct socket *sock; 2447 2448 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2449 if (sock != NULL) { 2450 err = __sys_shutdown_sock(sock, how); 2451 fput_light(sock->file, fput_needed); 2452 } 2453 return err; 2454 } 2455 2456 SYSCALL_DEFINE2(shutdown, int, fd, int, how) 2457 { 2458 return __sys_shutdown(fd, how); 2459 } 2460 2461 /* A couple of helpful macros for getting the address of the 32/64 bit 2462 * fields which are the same type (int / unsigned) on our platforms. 2463 */ 2464 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member) 2465 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen) 2466 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags) 2467 2468 struct used_address { 2469 struct sockaddr_storage name; 2470 unsigned int name_len; 2471 }; 2472 2473 int __copy_msghdr(struct msghdr *kmsg, 2474 struct user_msghdr *msg, 2475 struct sockaddr __user **save_addr) 2476 { 2477 ssize_t err; 2478 2479 kmsg->msg_control_is_user = true; 2480 kmsg->msg_get_inq = 0; 2481 kmsg->msg_control_user = msg->msg_control; 2482 kmsg->msg_controllen = msg->msg_controllen; 2483 kmsg->msg_flags = msg->msg_flags; 2484 2485 kmsg->msg_namelen = msg->msg_namelen; 2486 if (!msg->msg_name) 2487 kmsg->msg_namelen = 0; 2488 2489 if (kmsg->msg_namelen < 0) 2490 return -EINVAL; 2491 2492 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage)) 2493 kmsg->msg_namelen = sizeof(struct sockaddr_storage); 2494 2495 if (save_addr) 2496 *save_addr = msg->msg_name; 2497 2498 if (msg->msg_name && kmsg->msg_namelen) { 2499 if (!save_addr) { 2500 err = move_addr_to_kernel(msg->msg_name, 2501 kmsg->msg_namelen, 2502 kmsg->msg_name); 2503 if (err < 0) 2504 return err; 2505 } 2506 } else { 2507 kmsg->msg_name = NULL; 2508 kmsg->msg_namelen = 0; 2509 } 2510 2511 if (msg->msg_iovlen > UIO_MAXIOV) 2512 return -EMSGSIZE; 2513 2514 kmsg->msg_iocb = NULL; 2515 kmsg->msg_ubuf = NULL; 2516 return 0; 2517 } 2518 2519 static int copy_msghdr_from_user(struct msghdr *kmsg, 2520 struct user_msghdr __user *umsg, 2521 struct sockaddr __user **save_addr, 2522 struct iovec **iov) 2523 { 2524 struct user_msghdr msg; 2525 ssize_t err; 2526 2527 if (copy_from_user(&msg, umsg, sizeof(*umsg))) 2528 return -EFAULT; 2529 2530 err = __copy_msghdr(kmsg, &msg, save_addr); 2531 if (err) 2532 return err; 2533 2534 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE, 2535 msg.msg_iov, msg.msg_iovlen, 2536 UIO_FASTIOV, iov, &kmsg->msg_iter); 2537 return err < 0 ? err : 0; 2538 } 2539 2540 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys, 2541 unsigned int flags, struct used_address *used_address, 2542 unsigned int allowed_msghdr_flags) 2543 { 2544 unsigned char ctl[sizeof(struct cmsghdr) + 20] 2545 __aligned(sizeof(__kernel_size_t)); 2546 /* 20 is size of ipv6_pktinfo */ 2547 unsigned char *ctl_buf = ctl; 2548 int ctl_len; 2549 ssize_t err; 2550 2551 err = -ENOBUFS; 2552 2553 if (msg_sys->msg_controllen > INT_MAX) 2554 goto out; 2555 flags |= (msg_sys->msg_flags & allowed_msghdr_flags); 2556 ctl_len = msg_sys->msg_controllen; 2557 if ((MSG_CMSG_COMPAT & flags) && ctl_len) { 2558 err = 2559 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl, 2560 sizeof(ctl)); 2561 if (err) 2562 goto out; 2563 ctl_buf = msg_sys->msg_control; 2564 ctl_len = msg_sys->msg_controllen; 2565 } else if (ctl_len) { 2566 BUILD_BUG_ON(sizeof(struct cmsghdr) != 2567 CMSG_ALIGN(sizeof(struct cmsghdr))); 2568 if (ctl_len > sizeof(ctl)) { 2569 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL); 2570 if (ctl_buf == NULL) 2571 goto out; 2572 } 2573 err = -EFAULT; 2574 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len)) 2575 goto out_freectl; 2576 msg_sys->msg_control = ctl_buf; 2577 msg_sys->msg_control_is_user = false; 2578 } 2579 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS; 2580 msg_sys->msg_flags = flags; 2581 2582 if (sock->file->f_flags & O_NONBLOCK) 2583 msg_sys->msg_flags |= MSG_DONTWAIT; 2584 /* 2585 * If this is sendmmsg() and current destination address is same as 2586 * previously succeeded address, omit asking LSM's decision. 2587 * used_address->name_len is initialized to UINT_MAX so that the first 2588 * destination address never matches. 2589 */ 2590 if (used_address && msg_sys->msg_name && 2591 used_address->name_len == msg_sys->msg_namelen && 2592 !memcmp(&used_address->name, msg_sys->msg_name, 2593 used_address->name_len)) { 2594 err = sock_sendmsg_nosec(sock, msg_sys); 2595 goto out_freectl; 2596 } 2597 err = __sock_sendmsg(sock, msg_sys); 2598 /* 2599 * If this is sendmmsg() and sending to current destination address was 2600 * successful, remember it. 2601 */ 2602 if (used_address && err >= 0) { 2603 used_address->name_len = msg_sys->msg_namelen; 2604 if (msg_sys->msg_name) 2605 memcpy(&used_address->name, msg_sys->msg_name, 2606 used_address->name_len); 2607 } 2608 2609 out_freectl: 2610 if (ctl_buf != ctl) 2611 sock_kfree_s(sock->sk, ctl_buf, ctl_len); 2612 out: 2613 return err; 2614 } 2615 2616 static int sendmsg_copy_msghdr(struct msghdr *msg, 2617 struct user_msghdr __user *umsg, unsigned flags, 2618 struct iovec **iov) 2619 { 2620 int err; 2621 2622 if (flags & MSG_CMSG_COMPAT) { 2623 struct compat_msghdr __user *msg_compat; 2624 2625 msg_compat = (struct compat_msghdr __user *) umsg; 2626 err = get_compat_msghdr(msg, msg_compat, NULL, iov); 2627 } else { 2628 err = copy_msghdr_from_user(msg, umsg, NULL, iov); 2629 } 2630 if (err < 0) 2631 return err; 2632 2633 return 0; 2634 } 2635 2636 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg, 2637 struct msghdr *msg_sys, unsigned int flags, 2638 struct used_address *used_address, 2639 unsigned int allowed_msghdr_flags) 2640 { 2641 struct sockaddr_storage address; 2642 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2643 ssize_t err; 2644 2645 msg_sys->msg_name = &address; 2646 2647 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov); 2648 if (err < 0) 2649 return err; 2650 2651 err = ____sys_sendmsg(sock, msg_sys, flags, used_address, 2652 allowed_msghdr_flags); 2653 kfree(iov); 2654 return err; 2655 } 2656 2657 /* 2658 * BSD sendmsg interface 2659 */ 2660 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg, 2661 unsigned int flags) 2662 { 2663 return ____sys_sendmsg(sock, msg, flags, NULL, 0); 2664 } 2665 2666 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2667 bool forbid_cmsg_compat) 2668 { 2669 int fput_needed, err; 2670 struct msghdr msg_sys; 2671 struct socket *sock; 2672 2673 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2674 return -EINVAL; 2675 2676 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2677 if (!sock) 2678 goto out; 2679 2680 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0); 2681 2682 fput_light(sock->file, fput_needed); 2683 out: 2684 return err; 2685 } 2686 2687 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags) 2688 { 2689 return __sys_sendmsg(fd, msg, flags, true); 2690 } 2691 2692 /* 2693 * Linux sendmmsg interface 2694 */ 2695 2696 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen, 2697 unsigned int flags, bool forbid_cmsg_compat) 2698 { 2699 int fput_needed, err, datagrams; 2700 struct socket *sock; 2701 struct mmsghdr __user *entry; 2702 struct compat_mmsghdr __user *compat_entry; 2703 struct msghdr msg_sys; 2704 struct used_address used_address; 2705 unsigned int oflags = flags; 2706 2707 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2708 return -EINVAL; 2709 2710 if (vlen > UIO_MAXIOV) 2711 vlen = UIO_MAXIOV; 2712 2713 datagrams = 0; 2714 2715 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2716 if (!sock) 2717 return err; 2718 2719 used_address.name_len = UINT_MAX; 2720 entry = mmsg; 2721 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2722 err = 0; 2723 flags |= MSG_BATCH; 2724 2725 while (datagrams < vlen) { 2726 if (datagrams == vlen - 1) 2727 flags = oflags; 2728 2729 if (MSG_CMSG_COMPAT & flags) { 2730 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry, 2731 &msg_sys, flags, &used_address, MSG_EOR); 2732 if (err < 0) 2733 break; 2734 err = __put_user(err, &compat_entry->msg_len); 2735 ++compat_entry; 2736 } else { 2737 err = ___sys_sendmsg(sock, 2738 (struct user_msghdr __user *)entry, 2739 &msg_sys, flags, &used_address, MSG_EOR); 2740 if (err < 0) 2741 break; 2742 err = put_user(err, &entry->msg_len); 2743 ++entry; 2744 } 2745 2746 if (err) 2747 break; 2748 ++datagrams; 2749 if (msg_data_left(&msg_sys)) 2750 break; 2751 cond_resched(); 2752 } 2753 2754 fput_light(sock->file, fput_needed); 2755 2756 /* We only return an error if no datagrams were able to be sent */ 2757 if (datagrams != 0) 2758 return datagrams; 2759 2760 return err; 2761 } 2762 2763 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg, 2764 unsigned int, vlen, unsigned int, flags) 2765 { 2766 return __sys_sendmmsg(fd, mmsg, vlen, flags, true); 2767 } 2768 2769 static int recvmsg_copy_msghdr(struct msghdr *msg, 2770 struct user_msghdr __user *umsg, unsigned flags, 2771 struct sockaddr __user **uaddr, 2772 struct iovec **iov) 2773 { 2774 ssize_t err; 2775 2776 if (MSG_CMSG_COMPAT & flags) { 2777 struct compat_msghdr __user *msg_compat; 2778 2779 msg_compat = (struct compat_msghdr __user *) umsg; 2780 err = get_compat_msghdr(msg, msg_compat, uaddr, iov); 2781 } else { 2782 err = copy_msghdr_from_user(msg, umsg, uaddr, iov); 2783 } 2784 if (err < 0) 2785 return err; 2786 2787 return 0; 2788 } 2789 2790 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys, 2791 struct user_msghdr __user *msg, 2792 struct sockaddr __user *uaddr, 2793 unsigned int flags, int nosec) 2794 { 2795 struct compat_msghdr __user *msg_compat = 2796 (struct compat_msghdr __user *) msg; 2797 int __user *uaddr_len = COMPAT_NAMELEN(msg); 2798 struct sockaddr_storage addr; 2799 unsigned long cmsg_ptr; 2800 int len; 2801 ssize_t err; 2802 2803 msg_sys->msg_name = &addr; 2804 cmsg_ptr = (unsigned long)msg_sys->msg_control; 2805 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT); 2806 2807 /* We assume all kernel code knows the size of sockaddr_storage */ 2808 msg_sys->msg_namelen = 0; 2809 2810 if (sock->file->f_flags & O_NONBLOCK) 2811 flags |= MSG_DONTWAIT; 2812 2813 if (unlikely(nosec)) 2814 err = sock_recvmsg_nosec(sock, msg_sys, flags); 2815 else 2816 err = sock_recvmsg(sock, msg_sys, flags); 2817 2818 if (err < 0) 2819 goto out; 2820 len = err; 2821 2822 if (uaddr != NULL) { 2823 err = move_addr_to_user(&addr, 2824 msg_sys->msg_namelen, uaddr, 2825 uaddr_len); 2826 if (err < 0) 2827 goto out; 2828 } 2829 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT), 2830 COMPAT_FLAGS(msg)); 2831 if (err) 2832 goto out; 2833 if (MSG_CMSG_COMPAT & flags) 2834 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2835 &msg_compat->msg_controllen); 2836 else 2837 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr, 2838 &msg->msg_controllen); 2839 if (err) 2840 goto out; 2841 err = len; 2842 out: 2843 return err; 2844 } 2845 2846 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg, 2847 struct msghdr *msg_sys, unsigned int flags, int nosec) 2848 { 2849 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack; 2850 /* user mode address pointers */ 2851 struct sockaddr __user *uaddr; 2852 ssize_t err; 2853 2854 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov); 2855 if (err < 0) 2856 return err; 2857 2858 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec); 2859 kfree(iov); 2860 return err; 2861 } 2862 2863 /* 2864 * BSD recvmsg interface 2865 */ 2866 2867 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg, 2868 struct user_msghdr __user *umsg, 2869 struct sockaddr __user *uaddr, unsigned int flags) 2870 { 2871 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0); 2872 } 2873 2874 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags, 2875 bool forbid_cmsg_compat) 2876 { 2877 int fput_needed, err; 2878 struct msghdr msg_sys; 2879 struct socket *sock; 2880 2881 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT)) 2882 return -EINVAL; 2883 2884 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2885 if (!sock) 2886 goto out; 2887 2888 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0); 2889 2890 fput_light(sock->file, fput_needed); 2891 out: 2892 return err; 2893 } 2894 2895 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg, 2896 unsigned int, flags) 2897 { 2898 return __sys_recvmsg(fd, msg, flags, true); 2899 } 2900 2901 /* 2902 * Linux recvmmsg interface 2903 */ 2904 2905 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg, 2906 unsigned int vlen, unsigned int flags, 2907 struct timespec64 *timeout) 2908 { 2909 int fput_needed, err, datagrams; 2910 struct socket *sock; 2911 struct mmsghdr __user *entry; 2912 struct compat_mmsghdr __user *compat_entry; 2913 struct msghdr msg_sys; 2914 struct timespec64 end_time; 2915 struct timespec64 timeout64; 2916 2917 if (timeout && 2918 poll_select_set_timeout(&end_time, timeout->tv_sec, 2919 timeout->tv_nsec)) 2920 return -EINVAL; 2921 2922 datagrams = 0; 2923 2924 sock = sockfd_lookup_light(fd, &err, &fput_needed); 2925 if (!sock) 2926 return err; 2927 2928 if (likely(!(flags & MSG_ERRQUEUE))) { 2929 err = sock_error(sock->sk); 2930 if (err) { 2931 datagrams = err; 2932 goto out_put; 2933 } 2934 } 2935 2936 entry = mmsg; 2937 compat_entry = (struct compat_mmsghdr __user *)mmsg; 2938 2939 while (datagrams < vlen) { 2940 /* 2941 * No need to ask LSM for more than the first datagram. 2942 */ 2943 if (MSG_CMSG_COMPAT & flags) { 2944 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry, 2945 &msg_sys, flags & ~MSG_WAITFORONE, 2946 datagrams); 2947 if (err < 0) 2948 break; 2949 err = __put_user(err, &compat_entry->msg_len); 2950 ++compat_entry; 2951 } else { 2952 err = ___sys_recvmsg(sock, 2953 (struct user_msghdr __user *)entry, 2954 &msg_sys, flags & ~MSG_WAITFORONE, 2955 datagrams); 2956 if (err < 0) 2957 break; 2958 err = put_user(err, &entry->msg_len); 2959 ++entry; 2960 } 2961 2962 if (err) 2963 break; 2964 ++datagrams; 2965 2966 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */ 2967 if (flags & MSG_WAITFORONE) 2968 flags |= MSG_DONTWAIT; 2969 2970 if (timeout) { 2971 ktime_get_ts64(&timeout64); 2972 *timeout = timespec64_sub(end_time, timeout64); 2973 if (timeout->tv_sec < 0) { 2974 timeout->tv_sec = timeout->tv_nsec = 0; 2975 break; 2976 } 2977 2978 /* Timeout, return less than vlen datagrams */ 2979 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0) 2980 break; 2981 } 2982 2983 /* Out of band data, return right away */ 2984 if (msg_sys.msg_flags & MSG_OOB) 2985 break; 2986 cond_resched(); 2987 } 2988 2989 if (err == 0) 2990 goto out_put; 2991 2992 if (datagrams == 0) { 2993 datagrams = err; 2994 goto out_put; 2995 } 2996 2997 /* 2998 * We may return less entries than requested (vlen) if the 2999 * sock is non block and there aren't enough datagrams... 3000 */ 3001 if (err != -EAGAIN) { 3002 /* 3003 * ... or if recvmsg returns an error after we 3004 * received some datagrams, where we record the 3005 * error to return on the next call or if the 3006 * app asks about it using getsockopt(SO_ERROR). 3007 */ 3008 WRITE_ONCE(sock->sk->sk_err, -err); 3009 } 3010 out_put: 3011 fput_light(sock->file, fput_needed); 3012 3013 return datagrams; 3014 } 3015 3016 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, 3017 unsigned int vlen, unsigned int flags, 3018 struct __kernel_timespec __user *timeout, 3019 struct old_timespec32 __user *timeout32) 3020 { 3021 int datagrams; 3022 struct timespec64 timeout_sys; 3023 3024 if (timeout && get_timespec64(&timeout_sys, timeout)) 3025 return -EFAULT; 3026 3027 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32)) 3028 return -EFAULT; 3029 3030 if (!timeout && !timeout32) 3031 return do_recvmmsg(fd, mmsg, vlen, flags, NULL); 3032 3033 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys); 3034 3035 if (datagrams <= 0) 3036 return datagrams; 3037 3038 if (timeout && put_timespec64(&timeout_sys, timeout)) 3039 datagrams = -EFAULT; 3040 3041 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32)) 3042 datagrams = -EFAULT; 3043 3044 return datagrams; 3045 } 3046 3047 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg, 3048 unsigned int, vlen, unsigned int, flags, 3049 struct __kernel_timespec __user *, timeout) 3050 { 3051 if (flags & MSG_CMSG_COMPAT) 3052 return -EINVAL; 3053 3054 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL); 3055 } 3056 3057 #ifdef CONFIG_COMPAT_32BIT_TIME 3058 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg, 3059 unsigned int, vlen, unsigned int, flags, 3060 struct old_timespec32 __user *, timeout) 3061 { 3062 if (flags & MSG_CMSG_COMPAT) 3063 return -EINVAL; 3064 3065 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout); 3066 } 3067 #endif 3068 3069 #ifdef __ARCH_WANT_SYS_SOCKETCALL 3070 /* Argument list sizes for sys_socketcall */ 3071 #define AL(x) ((x) * sizeof(unsigned long)) 3072 static const unsigned char nargs[21] = { 3073 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3), 3074 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6), 3075 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3), 3076 AL(4), AL(5), AL(4) 3077 }; 3078 3079 #undef AL 3080 3081 /* 3082 * System call vectors. 3083 * 3084 * Argument checking cleaned up. Saved 20% in size. 3085 * This function doesn't need to set the kernel lock because 3086 * it is set by the callees. 3087 */ 3088 3089 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args) 3090 { 3091 unsigned long a[AUDITSC_ARGS]; 3092 unsigned long a0, a1; 3093 int err; 3094 unsigned int len; 3095 3096 if (call < 1 || call > SYS_SENDMMSG) 3097 return -EINVAL; 3098 call = array_index_nospec(call, SYS_SENDMMSG + 1); 3099 3100 len = nargs[call]; 3101 if (len > sizeof(a)) 3102 return -EINVAL; 3103 3104 /* copy_from_user should be SMP safe. */ 3105 if (copy_from_user(a, args, len)) 3106 return -EFAULT; 3107 3108 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a); 3109 if (err) 3110 return err; 3111 3112 a0 = a[0]; 3113 a1 = a[1]; 3114 3115 switch (call) { 3116 case SYS_SOCKET: 3117 err = __sys_socket(a0, a1, a[2]); 3118 break; 3119 case SYS_BIND: 3120 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]); 3121 break; 3122 case SYS_CONNECT: 3123 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]); 3124 break; 3125 case SYS_LISTEN: 3126 err = __sys_listen(a0, a1); 3127 break; 3128 case SYS_ACCEPT: 3129 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3130 (int __user *)a[2], 0); 3131 break; 3132 case SYS_GETSOCKNAME: 3133 err = 3134 __sys_getsockname(a0, (struct sockaddr __user *)a1, 3135 (int __user *)a[2]); 3136 break; 3137 case SYS_GETPEERNAME: 3138 err = 3139 __sys_getpeername(a0, (struct sockaddr __user *)a1, 3140 (int __user *)a[2]); 3141 break; 3142 case SYS_SOCKETPAIR: 3143 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]); 3144 break; 3145 case SYS_SEND: 3146 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3147 NULL, 0); 3148 break; 3149 case SYS_SENDTO: 3150 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3], 3151 (struct sockaddr __user *)a[4], a[5]); 3152 break; 3153 case SYS_RECV: 3154 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3155 NULL, NULL); 3156 break; 3157 case SYS_RECVFROM: 3158 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3], 3159 (struct sockaddr __user *)a[4], 3160 (int __user *)a[5]); 3161 break; 3162 case SYS_SHUTDOWN: 3163 err = __sys_shutdown(a0, a1); 3164 break; 3165 case SYS_SETSOCKOPT: 3166 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3], 3167 a[4]); 3168 break; 3169 case SYS_GETSOCKOPT: 3170 err = 3171 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3], 3172 (int __user *)a[4]); 3173 break; 3174 case SYS_SENDMSG: 3175 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1, 3176 a[2], true); 3177 break; 3178 case SYS_SENDMMSG: 3179 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], 3180 a[3], true); 3181 break; 3182 case SYS_RECVMSG: 3183 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1, 3184 a[2], true); 3185 break; 3186 case SYS_RECVMMSG: 3187 if (IS_ENABLED(CONFIG_64BIT)) 3188 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3189 a[2], a[3], 3190 (struct __kernel_timespec __user *)a[4], 3191 NULL); 3192 else 3193 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1, 3194 a[2], a[3], NULL, 3195 (struct old_timespec32 __user *)a[4]); 3196 break; 3197 case SYS_ACCEPT4: 3198 err = __sys_accept4(a0, (struct sockaddr __user *)a1, 3199 (int __user *)a[2], a[3]); 3200 break; 3201 default: 3202 err = -EINVAL; 3203 break; 3204 } 3205 return err; 3206 } 3207 3208 #endif /* __ARCH_WANT_SYS_SOCKETCALL */ 3209 3210 /** 3211 * sock_register - add a socket protocol handler 3212 * @ops: description of protocol 3213 * 3214 * This function is called by a protocol handler that wants to 3215 * advertise its address family, and have it linked into the 3216 * socket interface. The value ops->family corresponds to the 3217 * socket system call protocol family. 3218 */ 3219 int sock_register(const struct net_proto_family *ops) 3220 { 3221 int err; 3222 3223 if (ops->family >= NPROTO) { 3224 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO); 3225 return -ENOBUFS; 3226 } 3227 3228 spin_lock(&net_family_lock); 3229 if (rcu_dereference_protected(net_families[ops->family], 3230 lockdep_is_held(&net_family_lock))) 3231 err = -EEXIST; 3232 else { 3233 rcu_assign_pointer(net_families[ops->family], ops); 3234 err = 0; 3235 } 3236 spin_unlock(&net_family_lock); 3237 3238 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]); 3239 return err; 3240 } 3241 EXPORT_SYMBOL(sock_register); 3242 3243 /** 3244 * sock_unregister - remove a protocol handler 3245 * @family: protocol family to remove 3246 * 3247 * This function is called by a protocol handler that wants to 3248 * remove its address family, and have it unlinked from the 3249 * new socket creation. 3250 * 3251 * If protocol handler is a module, then it can use module reference 3252 * counts to protect against new references. If protocol handler is not 3253 * a module then it needs to provide its own protection in 3254 * the ops->create routine. 3255 */ 3256 void sock_unregister(int family) 3257 { 3258 BUG_ON(family < 0 || family >= NPROTO); 3259 3260 spin_lock(&net_family_lock); 3261 RCU_INIT_POINTER(net_families[family], NULL); 3262 spin_unlock(&net_family_lock); 3263 3264 synchronize_rcu(); 3265 3266 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]); 3267 } 3268 EXPORT_SYMBOL(sock_unregister); 3269 3270 bool sock_is_registered(int family) 3271 { 3272 return family < NPROTO && rcu_access_pointer(net_families[family]); 3273 } 3274 3275 static int __init sock_init(void) 3276 { 3277 int err; 3278 /* 3279 * Initialize the network sysctl infrastructure. 3280 */ 3281 err = net_sysctl_init(); 3282 if (err) 3283 goto out; 3284 3285 /* 3286 * Initialize skbuff SLAB cache 3287 */ 3288 skb_init(); 3289 3290 /* 3291 * Initialize the protocols module. 3292 */ 3293 3294 init_inodecache(); 3295 3296 err = register_filesystem(&sock_fs_type); 3297 if (err) 3298 goto out; 3299 sock_mnt = kern_mount(&sock_fs_type); 3300 if (IS_ERR(sock_mnt)) { 3301 err = PTR_ERR(sock_mnt); 3302 goto out_mount; 3303 } 3304 3305 /* The real protocol initialization is performed in later initcalls. 3306 */ 3307 3308 #ifdef CONFIG_NETFILTER 3309 err = netfilter_init(); 3310 if (err) 3311 goto out; 3312 #endif 3313 3314 ptp_classifier_init(); 3315 3316 out: 3317 return err; 3318 3319 out_mount: 3320 unregister_filesystem(&sock_fs_type); 3321 goto out; 3322 } 3323 3324 core_initcall(sock_init); /* early initcall */ 3325 3326 #ifdef CONFIG_PROC_FS 3327 void socket_seq_show(struct seq_file *seq) 3328 { 3329 seq_printf(seq, "sockets: used %d\n", 3330 sock_inuse_get(seq->private)); 3331 } 3332 #endif /* CONFIG_PROC_FS */ 3333 3334 /* Handle the fact that while struct ifreq has the same *layout* on 3335 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data, 3336 * which are handled elsewhere, it still has different *size* due to 3337 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit, 3338 * resulting in struct ifreq being 32 and 40 bytes respectively). 3339 * As a result, if the struct happens to be at the end of a page and 3340 * the next page isn't readable/writable, we get a fault. To prevent 3341 * that, copy back and forth to the full size. 3342 */ 3343 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg) 3344 { 3345 if (in_compat_syscall()) { 3346 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr; 3347 3348 memset(ifr, 0, sizeof(*ifr)); 3349 if (copy_from_user(ifr32, arg, sizeof(*ifr32))) 3350 return -EFAULT; 3351 3352 if (ifrdata) 3353 *ifrdata = compat_ptr(ifr32->ifr_data); 3354 3355 return 0; 3356 } 3357 3358 if (copy_from_user(ifr, arg, sizeof(*ifr))) 3359 return -EFAULT; 3360 3361 if (ifrdata) 3362 *ifrdata = ifr->ifr_data; 3363 3364 return 0; 3365 } 3366 EXPORT_SYMBOL(get_user_ifreq); 3367 3368 int put_user_ifreq(struct ifreq *ifr, void __user *arg) 3369 { 3370 size_t size = sizeof(*ifr); 3371 3372 if (in_compat_syscall()) 3373 size = sizeof(struct compat_ifreq); 3374 3375 if (copy_to_user(arg, ifr, size)) 3376 return -EFAULT; 3377 3378 return 0; 3379 } 3380 EXPORT_SYMBOL(put_user_ifreq); 3381 3382 #ifdef CONFIG_COMPAT 3383 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32) 3384 { 3385 compat_uptr_t uptr32; 3386 struct ifreq ifr; 3387 void __user *saved; 3388 int err; 3389 3390 if (get_user_ifreq(&ifr, NULL, uifr32)) 3391 return -EFAULT; 3392 3393 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu)) 3394 return -EFAULT; 3395 3396 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc; 3397 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32); 3398 3399 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL); 3400 if (!err) { 3401 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved; 3402 if (put_user_ifreq(&ifr, uifr32)) 3403 err = -EFAULT; 3404 } 3405 return err; 3406 } 3407 3408 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */ 3409 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd, 3410 struct compat_ifreq __user *u_ifreq32) 3411 { 3412 struct ifreq ifreq; 3413 void __user *data; 3414 3415 if (!is_socket_ioctl_cmd(cmd)) 3416 return -ENOTTY; 3417 if (get_user_ifreq(&ifreq, &data, u_ifreq32)) 3418 return -EFAULT; 3419 ifreq.ifr_data = data; 3420 3421 return dev_ioctl(net, cmd, &ifreq, data, NULL); 3422 } 3423 3424 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock, 3425 unsigned int cmd, unsigned long arg) 3426 { 3427 void __user *argp = compat_ptr(arg); 3428 struct sock *sk = sock->sk; 3429 struct net *net = sock_net(sk); 3430 const struct proto_ops *ops; 3431 3432 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) 3433 return sock_ioctl(file, cmd, (unsigned long)argp); 3434 3435 switch (cmd) { 3436 case SIOCWANDEV: 3437 return compat_siocwandev(net, argp); 3438 case SIOCGSTAMP_OLD: 3439 case SIOCGSTAMPNS_OLD: 3440 ops = READ_ONCE(sock->ops); 3441 if (!ops->gettstamp) 3442 return -ENOIOCTLCMD; 3443 return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD, 3444 !COMPAT_USE_64BIT_TIME); 3445 3446 case SIOCETHTOOL: 3447 case SIOCBONDSLAVEINFOQUERY: 3448 case SIOCBONDINFOQUERY: 3449 case SIOCSHWTSTAMP: 3450 case SIOCGHWTSTAMP: 3451 return compat_ifr_data_ioctl(net, cmd, argp); 3452 3453 case FIOSETOWN: 3454 case SIOCSPGRP: 3455 case FIOGETOWN: 3456 case SIOCGPGRP: 3457 case SIOCBRADDBR: 3458 case SIOCBRDELBR: 3459 case SIOCGIFVLAN: 3460 case SIOCSIFVLAN: 3461 case SIOCGSKNS: 3462 case SIOCGSTAMP_NEW: 3463 case SIOCGSTAMPNS_NEW: 3464 case SIOCGIFCONF: 3465 case SIOCSIFBR: 3466 case SIOCGIFBR: 3467 return sock_ioctl(file, cmd, arg); 3468 3469 case SIOCGIFFLAGS: 3470 case SIOCSIFFLAGS: 3471 case SIOCGIFMAP: 3472 case SIOCSIFMAP: 3473 case SIOCGIFMETRIC: 3474 case SIOCSIFMETRIC: 3475 case SIOCGIFMTU: 3476 case SIOCSIFMTU: 3477 case SIOCGIFMEM: 3478 case SIOCSIFMEM: 3479 case SIOCGIFHWADDR: 3480 case SIOCSIFHWADDR: 3481 case SIOCADDMULTI: 3482 case SIOCDELMULTI: 3483 case SIOCGIFINDEX: 3484 case SIOCGIFADDR: 3485 case SIOCSIFADDR: 3486 case SIOCSIFHWBROADCAST: 3487 case SIOCDIFADDR: 3488 case SIOCGIFBRDADDR: 3489 case SIOCSIFBRDADDR: 3490 case SIOCGIFDSTADDR: 3491 case SIOCSIFDSTADDR: 3492 case SIOCGIFNETMASK: 3493 case SIOCSIFNETMASK: 3494 case SIOCSIFPFLAGS: 3495 case SIOCGIFPFLAGS: 3496 case SIOCGIFTXQLEN: 3497 case SIOCSIFTXQLEN: 3498 case SIOCBRADDIF: 3499 case SIOCBRDELIF: 3500 case SIOCGIFNAME: 3501 case SIOCSIFNAME: 3502 case SIOCGMIIPHY: 3503 case SIOCGMIIREG: 3504 case SIOCSMIIREG: 3505 case SIOCBONDENSLAVE: 3506 case SIOCBONDRELEASE: 3507 case SIOCBONDSETHWADDR: 3508 case SIOCBONDCHANGEACTIVE: 3509 case SIOCSARP: 3510 case SIOCGARP: 3511 case SIOCDARP: 3512 case SIOCOUTQ: 3513 case SIOCOUTQNSD: 3514 case SIOCATMARK: 3515 return sock_do_ioctl(net, sock, cmd, arg); 3516 } 3517 3518 return -ENOIOCTLCMD; 3519 } 3520 3521 static long compat_sock_ioctl(struct file *file, unsigned int cmd, 3522 unsigned long arg) 3523 { 3524 struct socket *sock = file->private_data; 3525 const struct proto_ops *ops = READ_ONCE(sock->ops); 3526 int ret = -ENOIOCTLCMD; 3527 struct sock *sk; 3528 struct net *net; 3529 3530 sk = sock->sk; 3531 net = sock_net(sk); 3532 3533 if (ops->compat_ioctl) 3534 ret = ops->compat_ioctl(sock, cmd, arg); 3535 3536 if (ret == -ENOIOCTLCMD && 3537 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST)) 3538 ret = compat_wext_handle_ioctl(net, cmd, arg); 3539 3540 if (ret == -ENOIOCTLCMD) 3541 ret = compat_sock_ioctl_trans(file, sock, cmd, arg); 3542 3543 return ret; 3544 } 3545 #endif 3546 3547 /** 3548 * kernel_bind - bind an address to a socket (kernel space) 3549 * @sock: socket 3550 * @addr: address 3551 * @addrlen: length of address 3552 * 3553 * Returns 0 or an error. 3554 */ 3555 3556 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen) 3557 { 3558 struct sockaddr_storage address; 3559 3560 memcpy(&address, addr, addrlen); 3561 3562 return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address, 3563 addrlen); 3564 } 3565 EXPORT_SYMBOL(kernel_bind); 3566 3567 /** 3568 * kernel_listen - move socket to listening state (kernel space) 3569 * @sock: socket 3570 * @backlog: pending connections queue size 3571 * 3572 * Returns 0 or an error. 3573 */ 3574 3575 int kernel_listen(struct socket *sock, int backlog) 3576 { 3577 return READ_ONCE(sock->ops)->listen(sock, backlog); 3578 } 3579 EXPORT_SYMBOL(kernel_listen); 3580 3581 /** 3582 * kernel_accept - accept a connection (kernel space) 3583 * @sock: listening socket 3584 * @newsock: new connected socket 3585 * @flags: flags 3586 * 3587 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0. 3588 * If it fails, @newsock is guaranteed to be %NULL. 3589 * Returns 0 or an error. 3590 */ 3591 3592 int kernel_accept(struct socket *sock, struct socket **newsock, int flags) 3593 { 3594 struct sock *sk = sock->sk; 3595 const struct proto_ops *ops = READ_ONCE(sock->ops); 3596 struct proto_accept_arg arg = { 3597 .flags = flags, 3598 .kern = true, 3599 }; 3600 int err; 3601 3602 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, 3603 newsock); 3604 if (err < 0) 3605 goto done; 3606 3607 err = ops->accept(sock, *newsock, &arg); 3608 if (err < 0) { 3609 sock_release(*newsock); 3610 *newsock = NULL; 3611 goto done; 3612 } 3613 3614 (*newsock)->ops = ops; 3615 __module_get(ops->owner); 3616 3617 done: 3618 return err; 3619 } 3620 EXPORT_SYMBOL(kernel_accept); 3621 3622 /** 3623 * kernel_connect - connect a socket (kernel space) 3624 * @sock: socket 3625 * @addr: address 3626 * @addrlen: address length 3627 * @flags: flags (O_NONBLOCK, ...) 3628 * 3629 * For datagram sockets, @addr is the address to which datagrams are sent 3630 * by default, and the only address from which datagrams are received. 3631 * For stream sockets, attempts to connect to @addr. 3632 * Returns 0 or an error code. 3633 */ 3634 3635 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen, 3636 int flags) 3637 { 3638 struct sockaddr_storage address; 3639 3640 memcpy(&address, addr, addrlen); 3641 3642 return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address, 3643 addrlen, flags); 3644 } 3645 EXPORT_SYMBOL(kernel_connect); 3646 3647 /** 3648 * kernel_getsockname - get the address which the socket is bound (kernel space) 3649 * @sock: socket 3650 * @addr: address holder 3651 * 3652 * Fills the @addr pointer with the address which the socket is bound. 3653 * Returns the length of the address in bytes or an error code. 3654 */ 3655 3656 int kernel_getsockname(struct socket *sock, struct sockaddr *addr) 3657 { 3658 return READ_ONCE(sock->ops)->getname(sock, addr, 0); 3659 } 3660 EXPORT_SYMBOL(kernel_getsockname); 3661 3662 /** 3663 * kernel_getpeername - get the address which the socket is connected (kernel space) 3664 * @sock: socket 3665 * @addr: address holder 3666 * 3667 * Fills the @addr pointer with the address which the socket is connected. 3668 * Returns the length of the address in bytes or an error code. 3669 */ 3670 3671 int kernel_getpeername(struct socket *sock, struct sockaddr *addr) 3672 { 3673 return READ_ONCE(sock->ops)->getname(sock, addr, 1); 3674 } 3675 EXPORT_SYMBOL(kernel_getpeername); 3676 3677 /** 3678 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space) 3679 * @sock: socket 3680 * @how: connection part 3681 * 3682 * Returns 0 or an error. 3683 */ 3684 3685 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how) 3686 { 3687 return READ_ONCE(sock->ops)->shutdown(sock, how); 3688 } 3689 EXPORT_SYMBOL(kernel_sock_shutdown); 3690 3691 /** 3692 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket 3693 * @sk: socket 3694 * 3695 * This routine returns the IP overhead imposed by a socket i.e. 3696 * the length of the underlying IP header, depending on whether 3697 * this is an IPv4 or IPv6 socket and the length from IP options turned 3698 * on at the socket. Assumes that the caller has a lock on the socket. 3699 */ 3700 3701 u32 kernel_sock_ip_overhead(struct sock *sk) 3702 { 3703 struct inet_sock *inet; 3704 struct ip_options_rcu *opt; 3705 u32 overhead = 0; 3706 #if IS_ENABLED(CONFIG_IPV6) 3707 struct ipv6_pinfo *np; 3708 struct ipv6_txoptions *optv6 = NULL; 3709 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3710 3711 if (!sk) 3712 return overhead; 3713 3714 switch (sk->sk_family) { 3715 case AF_INET: 3716 inet = inet_sk(sk); 3717 overhead += sizeof(struct iphdr); 3718 opt = rcu_dereference_protected(inet->inet_opt, 3719 sock_owned_by_user(sk)); 3720 if (opt) 3721 overhead += opt->opt.optlen; 3722 return overhead; 3723 #if IS_ENABLED(CONFIG_IPV6) 3724 case AF_INET6: 3725 np = inet6_sk(sk); 3726 overhead += sizeof(struct ipv6hdr); 3727 if (np) 3728 optv6 = rcu_dereference_protected(np->opt, 3729 sock_owned_by_user(sk)); 3730 if (optv6) 3731 overhead += (optv6->opt_flen + optv6->opt_nflen); 3732 return overhead; 3733 #endif /* IS_ENABLED(CONFIG_IPV6) */ 3734 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */ 3735 return overhead; 3736 } 3737 } 3738 EXPORT_SYMBOL(kernel_sock_ip_overhead); 3739